Global case studies of social dialogue on AI and algorithmic management

Doellgast, Virginia; Appalla, Shruti; Ginzburg, Dina; Kim, Jeonghun; Thian, Wen Li Working Paper Global case studies of social dialogue on AI and algorithmic management ILO Working Paper, No. 144 Provided in Cooperation with: International Labour Organization (ILO), Geneva Suggested Citation: Doellgast, Virginia; Appalla, Shruti; Ginzburg, Dina; Kim, Jeonghun; Thian, Wen Li (2025) : Global case studies of social dialogue on AI and algorithmic management, ILO Working Paper, No. 144, ISBN 978-92-2-042168-0, International Labour Organization (ILO), Geneva, https://doi.org/10.54394/VOQE4924 This Version is available at: https://hdl.handle.net/10419/324259 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ XGlobal case studies of social dialogue on AI and algorithmic management Authors / Virginia Doellgast, Shruti Appalla, Dina Ginzburg, Jeonghun Kim, Wen Li Thian July / 2025 ILO Working Paper 144 © International Labour Organization 2025 Attribution 4.0 International (CC BY 4.0) This work is licensed under the Creative Commons Attribution 4.0 International. See: https:// creativecommons.org/licenses/by/4.0/. The user is allowed to reuse, share (copy and redistribute), adapt (remix, transform and build upon the original work) as detailed in the licence. The user must clearly credit the ILO as the source of the material and indicate if changes were made to the original content. Use of the emblem, name and logo of the ILO is not permitted in connection with translations, adaptations or other derivative works. Attribution – The user must indicate if changes were made and must cite the work as follows: Doellgast, V., Appalla, S., Ginzburg, D., Kim, J., Thian, W. Global case studies of social dialogue on AI and algorithmic management. ILO Working Paper 144. Geneva: International Labour Office, 2025.© ILO. Translations – In case of a translation of this work, the following disclaimer must be added along with the attribution: This is a translation of a copyrighted work of the International Labour Organization (ILO). This translation has not been prepared, reviewed or endorsed by the ILO and should not be considered an official ILO translation. The ILO disclaims all responsibility for its content and accuracy. Responsibility rests solely with the author(s) of the translation. Adaptations – In case of an adaptation of this work, the following disclaimer must be added along with the attribution: This is an adaptation of a copyrighted work of the International Labour Organization (ILO). This adaptation has not been prepared, reviewed or endorsed by the ILO and should not be considered an official ILO adaptation. The ILO disclaims all responsibility for its content and accuracy. Responsibility rests solely with the author(s) of the adaptation. Third-party materials – This Creative Commons licence does not apply to non-ILO copyright materials included in this publication. If the material is attributed to a third party, the user of such material is solely responsible for clearing the rights with the rights holder and for any claims of infringement. Any dispute arising under this licence that cannot be settled amicably shall be referred to arbitration in accordance with the Arbitration Rules of the United Nations Commission on International Trade Law (UNCITRAL). The parties shall be bound by any arbitration award rendered as a result of such arbitration as the final adjudication of such a dispute. For details on rights and licensing, contact: [email protected]. For details on ILO publications and digital products, visit: www.ilo.org/publns. ISBN 9789220421673 (print), ISBN 9789220421680 (web PDF), ISBN 9789220421697 (epub), ISBN 9789220421703 (html). ISSN 2708-3438 (print), ISSN 2708-3446 (digital) https://doi.org/10.54394/VOQE4924 The designations employed in ILO publications, which are in conformity with United Nations practice, and the presentation of material therein do not imply the expression of any opinion whatsoever on the part of the ILO concerning the legal status of any country, area or territory or of its authorities, or concerning the delimitation of its frontiers or boundaries. See: www.ilo. org/disclaimer. The opinions and views expressed in this publication are those of the author(s) and do not necessarily reflect the opinions, views or policies of the ILO. Reference to names of firms and commercial products and processes does not imply their endorsement by the ILO, and any failure to mention a particular firm, commercial product or process is not a sign of disapproval. Information on ILO publications and digital products can be found at: www.ilo.org/researchand-publications ILO Working Papers summarize the results of ILO research in progress, and seek to stimulate discussion of a range of issues related to the world of work. Comments on this ILO Working Paper are welcome and can be sent to resear[email protected]. Authorization for publication: Caroline Fredrickson, Director, Research Department ILO Working Papers can be found at: www.ilo.org/global/publications/working-papers Suggested citation: Doellgast, V., Appalla, S., Ginzburg, D., Kim, J., Thian, W. 2025. Global case studies of social dialogue on AI and algorithmic management, ILO Working Paper 144 (Geneva, ILO). https://doi. org/10.54394/VOQE4924 01 ILO Working Paper 144 Abstract Employers are adopting and refining artificial intelligence (AI) and algorithm-based tools in the workplace, with wide-ranging implications for work and employment. This working paper examines case studies of social dialogue on AI at national, regional, sectoral, company, and workplace levels in Europe, North America, Asia, South America and the Caribbean, and Africa. Findings are organized around three distinct ‘action fields’ in which worker representatives have sought to influence strategies and outcomes associated with the growing use of AI and algorithms in the workplace. These include the employment and skill impacts of AI, algorithmic management practices, and working conditions and rights in AI value chains. Across these action fields, social dialogue is playing a crucial role in encouraging an alternative, high road approach to AI investments and uses, based on complementing rather than replacing worker skills, empowering rather than controlling the workforce, and embedding rather than displacing new jobs in labor and social protections. Comparative findings suggest that these social dialogue initiatives are more effective where there are constraints on employer exit, support for collective worker voice, and strategies of inclusive labor solidarity. Key words: AI, algorithms, algorithmic management, social dialogue, labor unions, skills, job quality, surveillance, global value chains, outsourcing About the authors Virginia Doellgast is the Anne Evans Estabrook Professor of Employment Relations and Dispute Resolution in the ILR School at Cornell University. She is currently President of the Society for the Advancement of Socio-Economics (SASE), a Senior Research Fellow at the at the WSI-Hans Böckler Stiftung, and Co-Editor of the ILR Review. Her research focuses on the comparative political economy of labor markets and labor unions, inequality, precarity, and democracy at work. Publications include Exit, Voice, and Solidarity (Oxford University Press, 2022), Disintegrating Democracy at Work (Cornell University Press, 2012), International and Comparative Employment Relations (Sage, 2021), and Reconstructing Solidarity (Oxford University Press, 2018). Shruti Appalla is a PhD student in the ILR School at Cornell University. She studies the impact of emerging technologies on workers and firms in global supply chains. She has previously been a Predoctoral Fellow in Economics and holds a master’s degree in public policy from National Law School, India. Dina Ginzburg is an MS Labor Research and Policy student in the ILR School at Cornell University. She studies labor policy, union strategy, and collective action. Jeonghun Kim is a PhD student in the ILR School at Cornell University. His research explores how precarity is generated through new work arrangements and technologies, and how workers build solidarity to regulate them. He has conducted a comparative study of two unions’ organizing strategies in the South Korean food delivery platform sector. For his dissertation, he is working on a project that compares call center unions in the public health insurance sector in South Korea and the United States, focusing on how they respond to the challenges posed by AI adoption. 02 ILO Working Paper 144 Wen Li Thian is a PhD student in the ILR School at Cornell University. Her research focuses on mechanisms of labor control, the labor process, and technologies at work. She has done research on the lived experiences of factory workers and platform gig workers in Singapore. She holds a master’s degree in Sociology from the National University of Singapore. 03 ILO Working Paper 144 Abstract 01 About the authors 01 XIntroduction 07 Analytical framework 08 Case selection and research approach 12 X1 International, national, and regional social dialogue 14 1.1. Europe 14 1.1.1. Social dialogue at EU-level 14 1.1.2. Social dialogue at national level 17 1.2. North America 20 1.3. Asia 24 1.4. South America and the Caribbean 26 1.5. Africa 28 1.6. Summary 28 X2 Sectoral, company, and workplace social dialogue 31 2.1. Social dialogue over employment and skill impacts of AI: from labor replacing to labor complementing 31 2.1.1. Europe 32 2.1.2. North America 36 2.1.3. Asia, South America, and Africa 42 2.2. Social dialogue over algorithmic management: From labor controlling to labor empowering 45 2.2.1. Europe 46 2.2.2. North America 51 2.2.3. Asia, South America, the Caribbean, and Africa 54 2.3. Social dialogue over working conditions and rights in AI-enabled fissuring: From labor displacing to labor embedding 57 2.3.1. Re-embedding the AI value chain 58 2.4. Summary 64 XConclusion 66 Annex 1. List of interviews and email communication 69 References 71 Table of contents 04 ILO Working Paper 144 Acknowledgements 87 05 ILO Working Paper 144 List of Figures Figure 1: Supporting social dialogue on AI through constraints on exit, support for voice, and strategies of solidarity 11 Figure 1: Supporting social dialogue on AI through constraints on exit, support for voice, and strategies of solidarity 66 12 ILO Working Paper 144 bias have a disproportionate impact on women and minority ethnic groups.11 Successful social dialogue in this area thus depends on a broad and inclusive approach that represents the interests and concerns of the most vulnerable workers in a company, industry, or society. Finally, social dialogue that targets the social ‘embedding’ of jobs that have been created or restructured through labor-displacing AI innovations is likely to be most successful where it strengthens constraints on employer exit through strategies of inclusive labor solidarity. Efforts to improve conditions for workers across the AI supply chain have concentrated on strengthening basic employment rights and protections that make it more difficult to treat workers at the bottom of that chain as a precarious and disposable workforce. Social dialogue is strengthened where these workers act in solidarity with established labor unions and other organizations that can draw on more established institutional or labor market power. Our analysis of case studies below illustrates a wide variety of approaches to social dialogue in each of these action areas, which draw on different combinations of resources, but also face varied constraints. Overall, we argue these efforts are most successful in encouraging a more socially equitable and sustainable approach to AI adoption and deployment where labor unions are able to draw on stronger constraints on employer exit, support for collective worker voice, and strategies of inclusive labor solidarity. We return to this framework in our conclusion, to structure our discussion of the comparative findings. Case selection and research approach National, industry, and company cases and interviewees for this report were identified through conferences and meetings focused on social dialogue over AI, as well as through snowball sampling, via recommendations from contacts working on these topics in industry and policy roles. We provide more in-depth analysis on a subset of national cases, where we conducted multiple interviews with worker representatives. These include: Germany, the United States, South Korea, and India. In other cases, we conducted or draw on a smaller number of interviews or email communications, including France, Spain, Sweden, Canada, Japan, Brazil, the Dominican Republic, and Kenya. In Brazil, we interviewed representatives of employer associations. Examples from other countries discussed in the report are based primarily on archival sources. We also conducted interviews or email communications with academics with expertise in social dialogue over AI, and with union representatives at the global union federation UNI Global Union,12 which provided context and filled out details on some of the cases. In total, we conducted 19 regular interviews and 7 email communications or email-based interviews between August 2024 and January 2025. 13 The authors of this report also all participated in a conference at Cornell University on AI and the Future of Work organized jointly with the Communications Workers of America (CWA), the AFL-CIO Tech Institute, and UNI Global Union. Around 30 union representatives and organizers attended from the US, Canada, Germany, Sweden, Spain, and the Dominican Republic. They presented and discussed their experiences organizing around and negotiating over AI and algorithms in different industry and national settings. This conference was organized jointly with an academic conference, and involved exchange between international researchers and unions. 11 Kim, S., Oh, P., & Lee, J. (2024). Algorithmic gender bias: investigating perceptions of discrimination in automated decision-making. Behaviour & Information Technology, 43(16), 4208-4221. Kordzadeh, N., & Ghasemaghaei, M. (2022). Algorithmic bias: review, synthesis, and future research directions. European Journal of Information Systems, 31(3), 388-409. 12 UNI Global Union represents over 20 million service workers worldwide, including in care, commerce, finance, gaming, graphical and packaging, ICT and related services, media, entertainment & arts, post & logistics, and property services. https://uniglobalunion.org/ 13 See Appendix 1 for details on our interviews 13 ILO Working Paper 144 Our knowledge of and insights on many of the case studies discussed in this report benefited from these presentations and discussions. Reports published by the UC Berkeley Labor Center, UNI Europa, the Friedrich Ebert Stiftung, and the OECD describe and analyze case studies of social dialogue, including collective bargaining, over AI and digitalization, and are recommended for more information on agreements or initiatives.14 In addition, two databases provide more detailed language from collective agreements addressing technologies and their workforce consequences: the Public Services International (PSI) Digital Bargaining Hub and the UNI Europa Database of AI and Algorithmic Management in Collective Bargaining Agreements.15 We now turn to our case study findings. 14 Kresge, L. (2020). Union Collective Bargaining Agreement Strategies in Response to Technology. UC Berkeley Labor Center. Kresge, L. (2023) Negotiating Workers’ Rights at the Frontier of Digital Workplace Technologies in 2023. Berkeley Labor Center Blog. Brunnerová, S., D. Ceccon, B. Holubová, M. Kahancová, K. Lukáčová, G. Medas (2024) Collective Bargaining Practices on AI and Algorithmic Management in European Services Sectors. UNI Europa and Friedrich Ebert Stiftung. Rolf, S. (2024) AI and Algorithmic Management in European Services Sectors: Prevalence, functions, and a guide for negotiators. UNI Europa and Friedrich Ebert Stiftung. Global Deal (2024), Social Dialogue and the Use of Artificial Intelligence in the Workplace. OECD (2023), OECD Employment Outlook 2023: Artificial Intelligence and the Labour Market, OECD Publishing, Paris, https://doi.org/10.1787/08785bba-en. 15 Digital Bargaining Hub - PSI - The global union federation of workers in public services, https://publicservices.international/ digital-bargaining-hub; A database of AI and algorithmic management in collective bargaining agreements - UNI Europa, https://www. uni-europa. org/news/a-database-of-ai-and-algorithmic-management-in-collective-bargaining-agreements/ 14 ILO Working Paper 144 X1 International, national, and regional social dialogue Union involvement in social dialogue on AI at national and international levels can take different forms. Framework agreements between employers and unions, often with the involvement of government bodies, have established guiding principles concerning the development and adoption of AI-based tools, their employment and skills impacts, psychosocial health impacts, data privacy rights, and consultation or bargaining rights. Labor unions also are involved in consultation bodies or committees that advise on laws and policies establishing clear guidelines and allocating resources to enforcement or innovation and investment. Another form of social dialogue at these levels is via organizing and lobbying, as labor unions build coalitions with other civil society groups, or with employers, to shape policy decisions. 1.1. Europe 1.1.1. Social dialogue at EU-level The most robust examples of international social dialogue on AI and algorithms can be found within the European Union. This is due to the EU’s institutionalized forums for tripartite negotiation and consultation, through both the framework agreements signed between social partners and the process of developing and approving EU-level legislation or directives. Chagny and Blanc observe that AI governance in Europe is based on three forms of regulation: by formal laws or directives (for example, the AI Act); through bodies aimed at standardizing market tools, such as the International Organization for Standardization (ISO) and European standardization organizations; and through soft law, or self-regulation through charters, manifestos, or ethics committees.16 The 2020 European Social Partners Framework Agreement on Digitalization is an example of social dialogue at the EU level, between the European Trade Union Confederation (ETUC) and the major cross-sectoral employers’ associations Business Europe, SME United, and CEEP, with the participation of the European Commission.17 The agreement addresses four areas: ‘digital skills and securing employment’, ‘modalities of connecting and disconnecting’, ‘artificial intelligence and guaranteeing the human in control principle’, and ‘respect of human dignity and surveillance’. Suggested measures include training funds, learning accounts, and competence development plans; as well as schemes such as short-time work. The agreement also recommends that national affiliates should respect a series of principles when deploying AI systems, including 16 Chagny, O., & Blanc, N. (2024) Social dialogue as a form of bottom-up governance for AI: the experience in France. In Artificial Intelligence, Labour and Society. A. Ponce del Castillo (Ed.). ETUI. 197-205. P.200. https://www.etui.org/sites/default/files/2024-03/ Artificial%20intelligence%2C%20labour%20and%20society_2024.pdf#page=199 17 European Social Partners Framework Agreement on Digitalisation. June 2020. https://www.etuc.org/system/files/document/file202006/Final%2022%2006%2020_Agreement%20on%20Digitalisation%202020.pdf Framework agreements are non-binding, and thus do not include formal mechanisms for ensuring compliance. However, they provide agreed principles and recommended measures that can be referenced in more formal collective negotiations or policy initiatives. The text of the Framework Agreement states: ' In the context of article 155 of the Treaty [on the functioning of the European Union (TFEU)], this autonomous European framework agreement commits the members of BusinessEurope, SMEunited, CEEP and ETUC (and the liaison committee EUROCADRES/ CEC) to promote and to implement tools and measures, where necessary at national, sectoral and/or enterprise levels, in accordance with the procedures and practices specific to management and labour in the Member States and in the countries of the European Economic Area. The signatory parties also invite their member organisations in candidate countries to implement this agreement.' p.13 15 ILO Working Paper 144 ‘human in control’, prevention of harm, and avoiding bias and discrimination through risk assessment, transparency, and fairness.18 Special provisions are called for where AI is used in human resource decisions and analysis, including workers’ rights to request human oversight and to contest the decision. In 2022, an agreement on digitalization for central governments was concluded by the European Federation of Public Service Unions (EPSU) and European Public Administration Employers (EUPAE) - again, with the participation of the European Commission.19 It includes rights to training, to telework, to disconnect, to personal data protection, and to health and safety, committing employers to conduct health risk assessments in consultation with labor unions. Similar to the 2020 digitalization agreement, it also encourages a ‘human-in-command’ approach to AI. Sectoral agreements between EU social partners address similar themes. A 2020 framework agreement in electricity between EPSU, IndustriAll, and Eurelectric includes commitments to joint actions on training and lifelong learning associated with digitalization, as well as strategies to prevent psychosocial risks.20 Also in 2020, the EU social partners in the telecom sector, UNI Europa ICTS and ETNO, signed a Joint Declaration on Artificial Intelligence, which outlines agreed principles concerning ethical AI and the need to prioritize investing in digital skills and training within the telecoms industry, in the tech and telecoms ecosystems, and through policy and government action.21 More recently, EU social partners in the banking sector issued a 2024 EU Joint Declaration on Employment Aspects of Artificial Intelligence, which states that social dialogue, including collective bargaining, is important for managing risks and opportunities associated with AI in the industry.22 These include AI’s employment and skill impacts, occupational safety, and protection of digital rights.23 The Declaration encourages social dialogue to develop and put in place joint actions to support job transition and to ensure reor up-skilling opportunities when job profiles are affected by the growing use of AI and other digital technologies. EU legislation and directives provide important tools for worker representatives in different European countries to have information and bargaining rights concerning new technologies, and have often been developed through or influenced by tripartite consultation. The 1989 Framework Directive on occupational safety and health (OSH) includes provisions requiring that workers or their representatives are consulted in ‘the planning and introduction of new technologies to the workplace’; and requires employers to conduct risk assessments prior to purchasing new systems.24 In addition, the 2018 General Data Protection Regulation (GDPR) includes requirements concerning the collection, storage, and use of personal data, and its provisions are enforced through national data protection authorities. Both areas of legislation have been important tools for worker representatives contesting the use of algorithmic management tools.25 Using GDPR provisions, union representatives can block employers from processing employees’ personal data 18 Voss, E., & Bertossa, D. (2022). Collective Bargaining and Digitalization: A Global Survey of Union Use of Collective Bargaining to Increase Worker Control over Digitalization. New England Journal of Public Policy, 34(1), 10. 19 EPSU (2022) EU Social Partners signed new agreement on digitalization for central government. https://www.epsu.org/article/eusocial-partners-signed-new-agreement-digitalisation-central-government 20 Voss and Bertossa (2022), ibid, p.15 21 The Telecom Social Dialogue Committee (2020) Joint Declaration on Artificial Intelligence. https://www.uni-europa.org/olduploads/2020/12/20201130_UE-ETNO-declaration-AI.pdf 22 Joint Declaration on Employment Aspects of Artificial Intelligence by the European Social Partners in the Banking Sector. 14 May 2024. https://www.uni-europa.org/wp-content/uploads/sites/3/2024/05/Joint-Declaration-on-Employment-Aspects-of-Artificial-Intelligence_ BankSD.pdf 23 Email communication, Massimo Mensi, UNI Global Union, December 9, 2024. 24 Cefaliello, A. and J. Popma. (2024) How can workers protect themselves against the risks of new technologies? HesaMag 29 (Winter): 14-17, pp.15-16. 25 Aloisi, A. (2024). Regulating algorithmic management at work in the European Union: Data protection, non-discrimination and collective rights. International Journal of Comparative Labour Law and Industrial Relations, 40(1). 16 ILO Working Paper 144 to analyze and predict their behavior, including their performance.26 The GDPR also includes the possibility for Member States to provide, by law or in collective agreements, more specific rules to ensure the protection of the rights and freedoms of employees with regard to the processing of personal data in the context of employment (Article 88). More recently, the EU’s 2024 AI Act and 2024 Platform Work Directive have strengthened worker rights associated with digitalization, AI, and algorithmic management. Ponce Del Castillo observes that both ‘are notable examples of successful trade union influence’ in the regulatory domain – with improvements in worker rights made through dialogue with and the advocacy of unions.27 Provisions in the AI Act support building workers’ AI literacy and involvement in development of codes of conduct and design and development of AI systems. The AI Act also requires AI providers (i.e. developers and sellers) to identify and analyze ‘known and foreseeable risks’, in cases where an AI system is considered high-risk – and then to communicate this information to workers and their representatives.28 These include algorithmic management applications, such as promotion, dismissal, task allocation, monitoring, and performance evaluation. Certain applications, such as sentiment analysis tools widely used in monitoring systems, are prohibited in workplace and educational settings. Labor unions such as UNI Europa29 and IndustriALL30 have raised concerns with the AI Act’s reliance on voluntary compliance, its neglect of general-purpose AI, and exemptions or loopholes concerning certain algorithmic management tools. The European Trade Union Confederation (ETUC) has called on the EU to develop a range of new regulations and protections for workers, including a dedicated directive on algorithmic systems at work that ensures human oversight of AI-driven decision-making, an AI liability directive to ensure that workers operating AI systems are not held liable when AI causes harm, strengthened regulation of global digital value chains ensuring fair labor norms and ethical sourcing practices in AI development, and copyright protec - tions for creative workers that ensure informed consent, transparency, and fair remuneration.31 The ETUC is also coordinating the work of national labor unions to develop standards to operationalize the legal obligations in the AI Act, including ‘establishing a risk management system, maintaining a data governance programme, drawing up technical documentation, keeping 26 Le Bonniec, T. (2024) Another Path for AI Regulation: Worker Unions and Data Protection Rights. Italian Labour Law e-journal, 17 (2), pp.115-131. 10.6092/issn.1561-8048/20870. p.126. According to the text of the GDPR -https://eur-lex.europa.eu/eli/reg/2016/679/oj/eng - Analyzing and predicting employee behavior through data processing is considered ‘profiling’, and requires additional justification and transparency (Article 4). Union representatives can potentially challenge and potentially block employers from processing employee personal data to analyze and predict their behavior, including performance, by leveraging the 'right to object' provision (Article 21), which allows individuals to oppose processing of their data when it involves ‘profiling’, especially if such processing is considered excessive or intrusive, and could be seen as disproportionate to the stated business purpose. Article 22 gives the ‘data subject’ ‘the right not to be subject to a decision based solely on automated processing, including profiling, which produces legal effects concerning him or her or similarly significantly affects him or her.’ 27 Ponce Del Castillo, A. (2024) AI and trade unions: from rapid responses to proactive strategies. HesaMag 29 (Winter): 31-34. Pp. 3233. 28 The AI Act relies on providers of high-risk AI systems to self-assess and self-certify their products’ compliance, which has been criticized for not allowing for sufficient oversight by external bodies such as labor unions. Özkiziltan, D. and F. Landini (2025) Trustworthy and Human-Centric? The New Governance of Workplace AI Technologies under the EU’s Artificial Intelligence Act. Transfer, Issue 4. 29 UNI Europa (2021) AI Act - Our submission to the EU Commission. June 30, 2021. https://www.uni-europa.org/news/ai-act-our-submissionto-the-eu-commission/ 30 IndustriALL (2024) IndustriALL adopts a trade union strategy to tackle AI at work. December 4, 2024.https://news.industriall-europe. eu/Article/1176 31 ETUC (2025) Artificial Intelligence for Workers, Not Just for Profit: Ensuring Quality Jobs in the Digital Age. Adopted at the Executive Committee meeting of 04-05 March 2025. https://etuc.org/en/document/artificial-intelligence-workers-not-just-profit-ensuring-qualityjobs-digital-age 17 ILO Working Paper 144 automatic logs, providing instructions for use and being transparent to downstream deployers and users, and ensuring human oversight, robustness and cyber-security’.32 The Platform Work Directive similarly targets algorithmic transparency, empowering workers by requiring platforms33 to provide information to workers, their representatives, and ‘competent national authorities’ concerning how and why they are using algorithmic management systems (as well as their limitations). 1.1.2. Social dialogue at national level There are a number of examples of national-level social dialogue in Europe that target similar issues. The Nordic countries have a wide range of such initiatives.34 In Denmark, government-led commissions on digitalization and AI topics with tripartite representation have included the Disruption Council (2017), the Data Ethics Council (2019), the Sharing Economy Council (2019), and the Digitalization Partnership (2021).35 The Swedish government similarly established a tripartite Digitalization Commission in 2012, which was replaced by a Digitalization Council in 2017. The Swedish prime minister established an AI commission in 2023 with business, academic, media, and union representatives - with a focus on ‘assuring AI development that fosters common good’ and supporting social mobility and reskilling.36 In both countries, unions and employers have also been involved in consultations over legislation relating to AI and digitalization, including the recent AI Act. In Germany, the government’s AI Strategy was developed through expert bodies that included employer and labor union representatives, such as the German Bundestag’s 2018 Commission ‘Artificial Intelligence – Social Responsibility and Economic, Social and Ecological Potentials’. 37 This Strategy includes a government pledge ‘to put equal emphasis on the workers’ and companies’ interests, highlighting the importance of skills development, social security, health and safety, societal participation, and co-determination for the workers in the transformation process’.38 It has been implemented through, for example, a 2019 National Skills Strategy that supports investment in AI-specific skills; the Observatory for Artificial Intelligence in the World of Work to conduct research and inform policy; and the Civic Innovation Platform, which seeks to include representatives from civil society in AI application development ‘for social good’.39 German unions have also been involved in tripartite initiatives focused on ‘AI standardization’ to ‘develop standards for AI data models, security, criticality and quality criteria’.40 A working group on algorithmic management (AM) was established in 2023 by the German Federal Ministry of Labor and Social Affairs (BMAS) and the metal workers’ union IG Metall.41 This established four 32 Ponce Del Castillo (2024), Ibid, p.33. 33 A digital platform is a software based infrastructure that allows users to interact and conduct transactions over the internet. Platform work is ‘a form of employment in which organisations or individuals use an online platform to access other organisations or individuals to solve specific problems, or to provide specific services in exchange for payment. ’ European Council (2025) EU rules on platform work. https://www.consilium.europa.eu/en/policies/platform-work-eu/ 34 Ilsøe, A., Larsen, T. P., Mathieu, C., & Rolandsson, B. (2024). Negotiating about algorithms: Social partner responses to AI in Denmark and Sweden. ILR Review, 77(5), 856-868. 35 Ilsøe et al. (2024) Ibid. p. 859. 36 Ilsøe et al. (2024) Ibid. p. 863. For the final report of the AI commission, see: Regeringskansliet (2024) AI-kommissionens Färdplan för Sverige https://www.regeringen.se/rapporter/2024/11/ai-kommissionens-fardplan-for-sverige/ 37 Krzywdzinski, M., Gerst, D., & Butollo, F. (2023). Promoting human-centred AI in the workplace. Trade unions and their strategies for regulating the use of AI in Germany. Transfer: European Review of Labour and Research, 29(1), 53-70. Pp.59-60. 38 Özkiziltan, D. (2024). Governing Engels’ pause: AI and the world of work in Germany. ILR Review, 77(5), 846-856. p.849. 39 Özkiziltan, D. (2024). Ibid, p. 850. 40 Krzywdzinski et al. (2023), Ibid, p. 60. 41 Wotschack, P., Butollo, F., & Hellbach, L. (2024). Algorithmic management and democracy at work in Germany. Incoding Policy Brief, 2024. https://ddd.uab.cat/record/290693 18 ILO Working Paper 144 main action areas: 1) integrated AM system planning with worker participation; 2) transparency on data and function of AM systems; 3) building knowledge for assessing effects on work processes and conditions; and 4) knowledge-based change management with goal setting, evaluation, and feedback.42 A particularly noteworthy legislative development was the 2021 Works Council Modernization Act, which introduced AI-specific revisions to the ‘Works Constitution Act’ - the major legislation regulating works councils’ participation rights at company and workplace level.43 New provisions extend works councils’ consultation rights on new technologies to plans to adopt AI, and clarify that co-determination rights over selection guidelines for hiring, transfers and terminations include situations in which AI is used. Companies are also required to fund an expert (engaged by the works council) to consult on proposed changes or policies involving AI. In France, the 2022 French Digital Council (CNNum) involved consultation with labor unions, employers, citizens, and representatives from business, research, and government.44 The commission developed ‘roadmaps’ in three areas: combating online violence, digital transitions at work, and digital inclusion, with recommended legislation in each area. The 2024 national AI Commission included a representative from the French Democratic Confederation of Labor (CFDT). The commission developed recommendations that included the role of social dialogue in supervising digital transformation at work, investments in training, evaluating algorithms, and increasing environmental transparency.45 Tripartite dialogue in France is also important for translating EUlevel agreements and directives into national law, through national interprofessional agreements (ANI).46 For example, the European Agreement on Digitalization could be implemented through an ANI within France.47 One interesting project is the Social Dialogue on AI (DIAL-IA) initiative, which aims to both raise awareness of and resources supporting social dialogue at company and workplace levels in France.48 It is coordinated by the Institute of Social and Economic Research (IRES), a non-profit organization run by six labor unions, in collaboration with the National Agency for the Improvement of Working Conditions (ANACT). On the union side, it is led by CFDT, CFE-CGC, FO-Cadres and UGICT-CGT (later joined by CFTC). DIAL-IA brought together around 50 participants from unions, employers, and the public sector over 18 months. The project produced a joint manifesto for ‘technological social dialogue49 and a platform with tools to implement this dialogue, launched in January 2025. 42 Arbeitsgruppe 'Algorithmisches Management' (2023): Arbeitspapier: Daten und Gute Arbeit – Algorithmisches Management im Fokus https://www.denkfabrik bmas.de/fileadmin/Downloads/Publikationen/barrierefrei_BMAS_DF_Mantel_Algorithmis ches_Management. pdf 43 Özkiziltan, D. (2024). Governing Engels’ pause: AI and the world of work in Germany. ILR Review, 77(5), 846-856.p.849. BMAS (Bundesministerium für Arbeit und Soziales) 2025.Betriebsrätemodernisierungsgesetz. https://www.bmas.de/DE/Service/Gesetze-und-Gesetzesvorhaben/betriebsraetemodernisierungsgesetz.html 44 Chagny, O., & Blanc, N. (2024) Social dialogue as a form of bottom-up governance for AI: the experience in France. Artificial Intelligence, Labour and Society. A. Ponce del Castillo (Ed.). ETUI. 197-205. https://www.etui.org/sites/default/files/2024-03/Artificial%20 intelligence%2C%20labour%20and%20society_2024.pdf#page=199 CNR Numérique - Méthode, enseignements et feuilles de route. https://cnnumerique.fr/annonce/cnr-numerique-methode-enseignements-et-feuilles-de-route 45 CFDT (2024) Commission de l'IA : Faire des travailleurs les acteurs de transformations numériques justes et responsables. https:// www.cfdt.fr/sinformer/communiques-de-presse/commission-de-lia-faire-des-travailleurs-les-acteurs-de-transformations-numeriquesjustes-et-responsables# . Commission de l’Intelligence Artificielle (2024) IA: Notre Ambition Pour La France. https://www.info.gouv. fr/upload/media/content/0001/09/4d3cc456dd2f5b9d79ee75feea63b47f10d75158.pdf 46 National interprofessional agreements (ANI) are tripartite agreements that are signed by the social partners, and cover different areas of employment and social policy. Most ANIs need to be transposed into legislation before they can be implemented. French law requires the government to organize a dialogue prior to introducing a bill to Parliament in certain areas of reform, although exceptions can be made for ‘urgent circumstances’. Vincent, C. (2019). France: The rush towards prioritising the enterprise level. In Collective Bargaining in Europe: Towards an Endgame. T. Müller, & J. Waddington (Eds.). Brussels: European Trade Union Institute. 217-238. 47 Interview, Franca Salis-Manidier, CFDT, January 13, 2025 48 DIAL IA Platform (2025) http://dial-ia.fr/ 49 DIAL-IA (2025) Pour un dialogue social au service des bons usages de l’IA et d’une nouvelle étape de progrès social dans les entreprises et les administrations. https://www.actuel-rh.fr/sites/default/files/article-files/dialia_manifeste.pdf 19 ILO Working Paper 144 In addition, European labor unions have been involved in framing national legislation. In Spain, the 2021 Riders Law (La Ley Rider) was developed through a tripartite agreement between the Workers’ Commissions (CC.OO.) and the General Union of Workers (UGT) and the Spanish Confederation of Employers’ Organizations (CEOE) and the Spanish Confederation of Small and Medium Enterprises (CEPYME). It was pioneering in recognizing platform-based food delivery courriers (“riders”) as employees rather than independent contractors (under certain conditions), and requiring platforms to disclose information to riders on how algorithms and AI are used in hiring decisions, layoffs, worker profiles; and their impacts on working conditions – with requirements that worker representatives are informed of the algorithm’s ‘parameters, rules, and instructions’.50 Tripartite negotiations that lasted 5 months in Spain also contributed to revising of the Workers’ Statute Law to incorporate the right to algorithmic transparency, which requires every type of platform to inform works councils about the inner workings of the algorithm ‘that may affect working conditions, access to and maintenance of employment, including the creation of profiles.’51 In other cases, national agreements have established key principles or guidelines for bargaining on digitalization at lower levels. In Spain, the 2023 Agreement for Employment and Collective Bargaining, negotiated between the unions CC.OO. and UGT and the employer organizations CEOE and CEPYME, includes requirements that sector and company agreements establish procedures for informing workers of digitalization projects and their impacts on employment, skills, and working conditions. It also includes a section on AI with clear requirements that AI systems ‘follow the principle of human control’ and that companies provide worker representatives with ‘transparent and understandable information on processes based on AI in human resources procedures (hiring, evaluation, promotion and dismissal) and ensure that there is no prejudice or discrimination.’52 Another example here is the Basic Agreement in Norway, negotiated at the national level between the Confederation of Norwegian Enterprise (NHO) and the Norwegian Confederation of Trade Unions (LO). It includes provisions to support worker privacy, bias prevention, and worker representative involvement in decision-making associated with AI tools in the workplace.53 The agreement also makes reference to national legislation, including the Working Environment Act (on the need to consult with shop stewards on changes) and the Data Privacy Act (on the collection and storage of personal data). While the UK is no longer in the EU, much of its legislation is formally harmonized with EU Directives and Laws, including the UK GDPR. The Trades Union Congress (TUC) set up its own taskforce and drafted the ‘AI Regulation and Employment Rights Bill’ in 2024, with the stated purpose to regulate ‘the use of artificial intelligence systems by employers in relation to workers, employees and jobseekers to protect their rights and interests in the workplace’.54 It includes provisions supporting union rights regarding employers’ use of AI systems and risk mitigation associated with AI value chains. A Special Advisory Committee provided input, including representatives from universities, research institutes, employer organizations, and labor unions, and 50 Eurofound (2021), Riders’ law(Initiative), Record number 2449, Platform Economy Database, Dublin, https://apps.eurofound.europa. eu/platformeconomydb/riders-law-105142. 51 European Agency for Safety and Health at Work. (2022). Spain: The ‘riders’ law’, new regulation on digital platform work. In OSHA Europa. https://osha.europa.eu/sites/default/files/2022-01/Spain_Riders_Law_new_regulation_digital_platform_work.pdf 52 BOE-A-2023-12870 Resolución de 19 de mayo de 2023, de la Dirección General de Trabajo, por la que se registra y publica el V Acuerdo para el Empleo y la Negociación Colectiva. (2023). Ministry of Labor and Social Economy. https://www.boe.es/diario_boe/ txt.php?id=BOE-A-2023-12870 53 Brunnerová et al. (2024), Ibid, p. 20. 54 TUC (2024) Artificial Intelligence (Regulation and Employment Rights) Bill. https://www.tuc.org.uk/research-analysis/reports/artificialintelligence-regulation-and-employment-rights-bill 20 ILO Working Paper 144 members of parliament. While the Bill was never formally introduced in parliament, its framing encouraged dialogue on these topics, establishing a common union position and leadership. In Wales, the tripartite Workforce Partnership Council, a forum for public services workforce issues, released a set of principles for ‘managing the transition to a digital workplace’ – including a focus on promoting ‘a social partnership approach to the involvement, participation and consultation of trade unions’.55 Following the recommendation of the Workforce Partnership Council, the Welsh Government adopted new guidelines on algorithmic management and workers rights.56 The union confederation TUC Cymru also worked in collaboration with the organization Connected by Data to develop a toolkit for empowering worker voice and public sector procurement of data and AI.57 1.2. North America Tripartite social dialogue is less institutionalized in North America compared to in Europe. However, there are many examples of union-led initiatives at national and regional or local (e.g. state, province, city) levels, in which labor unions have sought to formulate models and shape legislation on AI and algorithms. These are often organized in coalition with other civil society organizations or in consultation with employers and politicians or government representatives. In the United States, social dialogue at the national level involves listening sessions, meetings with policy-makers, and lobbying or organizing efforts focused on influencing the content of specific policies or their implementation through agencies. In 2023, the White House conducted a Listening Session with Union Leaders on Advancing Responsible Artificial Intelligence Innovation, which included representatives from the American Federation of Labor - Congress of Industrial Organization (AFL-CIO) Tech Institute, American Federation of State, County and Municipal Employees (AFSCME), American Federation of Teachers (AFT), Communications Workers of America (CWA), Screen Actors Guild–American Federation of Television and Radio Artists (SAGAFTRA), International Alliance of Theatrical Stage Employees (IATSE), International Brotherhood of Teamsters (IBT), National Education Association (NEA), United Auto Workers (UAW), United Food and Commercial Workers International Union (UFCW), and Writers Guild of America (WGA) East.58 In 2023, Vice President Harris convened a separate meeting of consumer protection, labor, and civil rights leaders to discuss AI-based risks.59 These constituted informal consultations with labor unions and civil society organizations, in a context where unions, business leaders, and technology developers were also providing written statements and advice. For example, the White House Office of Science and Technology Policy (OSTP) put out a Request for Information 55 Workforce Partnership Council agreement: partnership and managing change. 2 December 2021.https://www.gov.wales/workforcepartnership-council-agreement-partnership-and-managing-change-html 56 Welsh Government (2024) Managing technology that manages people: A Social Partnership approach to algorithmic management systems in the Welsh public sector. https://www.gov.wales/managing-technology-manages-people 57 Cantwell-Corn, A. (2025) Toolkit: Worker voice in public sector procurement of digital and AI systems in Wales. Feb 12, 2025. https:// connectedbydata.org/resources/worker-voice-procurement-guidance-wales 58 The White House (2023) Readout of White House Listening Session with Union Leaders on Advancing Responsible Artificial Intelligence Innovation. July 3, 2023. https://www.whitehouse.gov/briefing-room/statements-releases/2023/07/03/readout-of-white-house-listening-session-with-unionleaders-on-advancing-responsible-artificial-intelligence-innovation/ 59 The White House (2023) Readout of Vice President Harris’s Meeting with Consumer Protection, Labor, and Civil Rights Leaders on AI, July 13, 2023. https://www.whitehouse.gov/briefing-room/statements-releases/2023/07/13/readout-of-vice-president-harriss-meetingwith-consumer-protection-labor-and-civil-rights-leaders-on-ai/ 21 ILO Working Paper 144 on Automated Worker Surveillance and Management in May, 2023 – and unions responded with formal comments, intended to inform policy making in this area.60 Former US President Biden issued a 2023 Executive Order (EO) on the Safe, Secure, and Trustworthy Development and Use of AI, which affirmed the importance of social dialogue in AI adoption and deployment: ‘as AI creates new jobs and industries, all workers need a seat at the table, including through collective bargaining.’ Federal agencies were tasked with developing principles and best practices that mitigate worker harms and maximize worker benefits.61 President Trump repealed this and issued a new Executive Order on Removing Barriers to American Leadership in Artificial Intelligence, which directs White House Officials to develop an action plan to ‘promote human flourishing, economic competitiveness, and national security’62. In April 2025, the White House Office of Management and Budget issued two new memos aimed at providing guardrails concerning how federal agencies use and purchase AI, which include risk management practices where systems are deemed ‘high-risk’, with significant potential impacts.63 Under the Biden Administration, federal agencies, including the National Labor Relations Board (NLRB) and the Equal Employment Opportunity Commission (EEOC) also issued AI-specific guidance or opinions - for example, clarifying labor law protections against the use of surveillance tools by employers in union organizing campaigns, and the scope of liability by third-party AI software providers under anti-discrimination laws. Litwin and Racabi observe that these kinds of initiatives can be overturned by the courts, as well by a new administration.64 Indeed, these agencies had already rescinded a number of policies at the time of writing; for example, in February 2025, the NLRB Acting General Council rescinded at least 18 memoranda issued by his predecessor, including one addressing the impact of electronic monitoring on employee rights.65 Unlike Europe, the US does not have a federal data privacy law; and almost all of the 20 state-level data privacy laws explicitly exclude workers.66 However, unions have sought to encourage strengthened worker rights in draft bills, with some success. 67 One example is the 2018 California Consumer Protection Act (CCPA), which is unique in giving workers rights of information, access, and opt outs of employer data collection policies, as well as protection from retaliation 60 The report notes that 211 stakeholders submitted comments, including ‘91 workers, 19 advocacy organizations, 16 researchers and research organizations, 12 unions, 10 trade associations, eight technology developers, one coalition comprised of advocacy organizations and a union, and 54 unspecified stakeholders’. U.S. Government Accountability Office (2024) Digital Surveillance of Workers: Tools, Uses, and Stakeholder Perspectives. GAO-24-107639. GAO-24-107639, Digital Surveillance of Workers: Tools, Uses, and Stakeholder Perspectives 61 The White House (2023) Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence, October 30, 2023. https://www.whitehouse.gov/briefing-room/presidential-actions/2023/10/30/executive-order-on-the-safe-secure-and-trustworthydevelopment-and-use-of-artificial-intelligence/ Guidance to federal agencies on best practices included ‘centering worker empowerment’ (workers and their representatives should be informed of and have genuine input in the design, development, testing, training, use, and oversight of AI systems for use in the workplace), ethically developing AI, establishing AI governance and human oversight, ensuring transparency in AI use, protecting labor and employment rights, using AI to enable workers, supporting workers impacted by AI, and ensuring responsible use of worker data. Section 6 directed the Secretary of Labor to collaborate with labor unions to create best practices for employer use of AI, including addressing concerns with job displacement, protected activity, and data collection transparency. 62 The White House (2025) Removing Barriers to American Leadership in Artificial Intelligence. January 23, 2025. https://www.whitehouse. gov/presidential-actions/2025/01/removing-barriers-to-american-leadership-in-artificial-intelligence/ 63 Alder, M. (2025) Trump White House releases guidance for AI use, acquisition in government. FedScoop, April 4, 2025. https://fedscoop. com/trump-white-house-ai-use-acquisition-guidance-government/ 64 Litwin, A. S., & Racabi, G. (2024). Varieties of AI Regulations: The United States Perspective. ILR Review, 77(5), 799-812. p.809. 65 Stanek, T. M. and Z. V. Zagger (2025) NLRB Acting General Council Rescinds Many of Predecessor’s Memos, Sets Stage for New Labor Policy. The National Law Review, February 15, 2025. https://natlawreview.com/article/nlrb-acting-general-counsel-rescinds-manypredecessors-memos-sets-stage-new-labor 66 Khan, M., A. Bernhardt, L. Pathak. (2024) Current Landscape of Tech and Work Policy: A Roundup of Key Concepts. https://laborcenter. berkeley.edu/tech-and-work-policy-guide/ 67 Kang, C. (2024) States take up AI regulation amid federal standstill, New York Times, June 10, 2024.https://www.nytimes.com/2024/06/10/ technology/california-ai-regulation.html 28 ILO Working Paper 144 1.5. Africa In Africa, we were not able to identify social dialogue examples that directly involved labor unions. However, labor and employment topics have been included in policies and strategy documents. At the international level, the African Union – made up of 55 member states – published a ‘Continental Artificial Intelligence Strategy.117 This white paper mentions the need to identify and bridge regulatory gaps - including standards for the public procurement of AI systems, and regulatory approval of AI for use as medical devices within health systems. It also calls for assessing AI’s implications for the African labor market and impact on vulnerable groups, with goals of avoiding exacerbated socioeconomic inequalities and of developing a national policy to address labor transformations. Seven African nations (Benin, Egypt, Ghana, Mauritius, Rwanda, Senegal, and Tunisia) have drafted national AI strategies.118 In their analysis of regulation of labor in AI in East and Southern Africa, Bischoff et al. argue that most regulation in the region has focused on the support and development of AI-based industry.119 At the same time, they give examples of national legislation providing consumer and worker protections, including a Data Protection Act in Botswana, efforts to limit race-based credit profiling in Eswatini, and legislation promoting transparency in collection and use of individual data in Namibia and South Africa. In South Africa, a new industry association, the South African Artificial Intelligence Association (SAAIA), brings together representatives from commerce, government, NGOs, academia, and start-ups - but, as Bischoff et al. emphasize, labor unions are not included. It discusses ethical concerns in the areas of privacy and bias and seeks to encourage more responsible AI practices. In Kenya, with the introduction of the M-Pesa, a money transfer service, there has been a boom in the Fintech sector. This has also opened doors to further innovation and rapid infrastructural development around the use of AI to address concerns like food security, affordable housing, manufacturing, and affordable health care.120 Kenya has also become a popular location for sourcing content moderation and data labelling.121 Kenya passed a Data Protection Act in 2019, addressing concerns around data subject rights, privacy and data minimization.122 1.6. Summary Labor unions have been involved in a range of social dialogue initiatives to influence policy or develop framework agreements and principles on AI and algorithms. Our analysis above suggests two broad conclusions. First, union involvement in social dialogue at the national or international levels differs significantly across countries and world regions. Europe has the largest number of well-documented examples of tripartite committees and consultation bodies focusing on AI and digitalization 117 African Union (2024) Continental Artificial Intelligence Strategy. https://au.int/en/documents/20240809/continental-artificial-intelligencestrategy 118 Okolo, C. T. (2024, March 15). Reforming data regulation to advance AI governance in Africa. Brookings. https://www.brookings.edu/ articles/reforming-data-regulation-to-advance-ai-governance-in-africa/ 119 Bischoff, C., Kamoche, K., & Wood, G. (2024). The Formal and Informal Regulation of Labor in AI: The Experience of Eastern and Southern Africa. ILR Review, 77(5), 825-835. 120 Mgala, M.. 2020. The extent and use of Artificial Intelligence to achieve the Big Four Agenda in Kenya. Multidisciplinary Journal of Technical University of Mombasa 1(1):1–7 121 Graham, M., I. Hjorth, & V. Lehdonvirta. 2017. Digital labor and development: Impacts of global digital labor platforms and the gig economy on worker livelihoods. Transfer: European Review of Labor and Research 23(2):135–62. 122 Mary Kageni and Yvonne Odhiambo. Strengthening data protection in Kenya: Opportunities and the way forward – KIPPRA. (n.d.). Retrieved January 14, 2025, from https://kippra.or.ke/strengthening-data-protection-in-kenya-opportunities-and-the-way-forward/ 29 ILO Working Paper 144 topics, and of framework agreements signed between employers and unions on these issues. The agreements or recommendations of these bodies have fed directly into laws and policies at EU-level and within EU member states. South Korea is another case where there has been more formal, national-level social dialogue that has had direct influence on policies addressing AI, digitalization, and platform work. However, social dialogue also takes a range of forms that can be difficult to compare across countries. Labor unions are involved in shaping AI policy and standard-setting through more informal information sharing, consultation processes, and stakeholder meetings. They also influence policy through lobbying and campaigning efforts that develop model language for bills or regulations and that seek to mobilize members and the public in support of more labor-friendly provisions. In the US, for example, unions have worked in coalition with other groups to influence executive orders and legislation at state and local levels that strengthen data protection and government procurement policies relating to AI tools. Thus, an analysis of policies strengthening labor rights or establishing ethical principles can also provide comparative insight into the ‘collective voice’ or influence of labor unions. Several of our country cases, including India, the Dominican Republic, and Kenya, stand out for not yet having national AI or data protection policies that address labor and employment concerns. However, there have been efforts in these countries by labor organizations and NGOs to use leverage from existing laws to strengthen or extend employment protections to new or poorly protected worker groups. Many of these focus on the BPO industry, which includes both traditional call center and back office workers and newer data labelling and content moderation workers - with both groups often subjected to algorithmic surveillance. Second, we find some degree of cross-national convergence on common principles widely held to be ‘best practices’ for the ethical development and deployment of AI. These are repeated across framework agreements, policy statements, and executive orders. Their positions, summarized below, can be organized under the three main themes in our report: 1. From labor replacing to complementing: AI should contribute to enhancing and complementing worker skills, and workers should have discretion over how they use these tools in their jobs and workplaces. Employers should invest in digital skills while securing employment or sharing productivity gains with workers. As stated succinctly in a report by the German Trade Union Confederation (DGB): ‘The focus must be on expanding human capabilities through AI, and not replacing people with machines.’123 Laws and policies encouraging employers to adopt AI-based tools in a way that is ‘labor complementing’ include those that: ●strengthen employment or job security ●invest in training and digital skills ●include clear rules concerning uses of generative AI to reproduce art, writing, voice, or images that protect copyright and transparency, and that ensure fair compensation for creators. 2. From labor controlling to empowering: AI should be adopted based on a ‘human-in-command’ approach, with human decision-makers having oversight on how algorithm-based 123 DGB (2019) Künstliche Intelligenz und die Arbeit von morgen: Ein Impulspapier des Deutschen Gewerkschaftsbundes zur Debatte um Künstliche Intelligenz (KI) in der Arbeitswelt. Cited in:Özkiziltan, D. (2024) Ibid, p.851. 30 ILO Working Paper 144 management tools are adopted and used. Workers should have recourse to contest decisions made by algorithms that affect their work and employment, particularly where these decisions may be biased or based on opaque models. Workers should have rights to personal data protection, and to health and safety safeguards that include protections against psychosocial risks.124 Human dignity should be respected. Laws and policies encouraging employers to adopt AI-based tools in a way that is ‘labor empowering’ include those that: ●require human oversight of AI-based decisions ●prohibit uses of AI tools with significant risks of bias or psychosocial harm ●strengthen data protection rights ●develop AI-specific health and safety guidelines and protections 3. From labor displacing to embedding: AI-based technologies, including algorithmic management tools, should not be used by firms to expand worker precarity and undermine job quality through intensified subcontracting, relocation of jobs, or casualized contracts. AI-related labor in AI and digital value chains should be ethically sourced, and workers should be protected from exploitative conditions. Laws and policies that encourage the social ‘embedding’ of AI-enabled fissured labor include those that: ●extend employment rights and protections to subcontracted and temporary workers ●strengthen regulation of global AI and digital value chains ● support union organizing and bargaining rights for data and content moderation workers Most fundamentally, where labor unions are involved in social dialogue, they have sought to support provisions in policies and framework agreements that establish worker representatives’ information, consultation, and bargaining rights, concerning how AI is being used and its effects on work and workers. This includes the ability to carry out risk assessments, but also to have a substantive voice in decisions concerning how these technologies are used in the workplace. A union representative involved in a national social dialogue body explained that even where unions play more of a consultation role, their impact can be significant: ‘It is important when you are in such a commission to bring your point of view, because otherwise it could be a dead issue, nobody would be talking about social dialogue.’125 In the following section, we ask to what extent and in what ways worker representatives have been able to leverage different laws, policies, and framework or collective agreements to negotiate or consult over specific uses of AI and algorithms. How is social dialogue at the workplace, firm, and sector level being adapted to the new challenges of AI? 124 Work-related psychosocial risks or hazards are defined as ‘factors in the work environment that can cause stress, strain, or interpersonal problems for the worker’ and are a major contributor to workplace injury and disability. Schulte, P. et al. (2024) An urgent call to address work‐related psychosocial hazards and improve worker well‐being. American Journal of Industrial Medicine, 67(6), 499-514. 125 Interview, Franca Salis-Manidier, CFDT, January 13, 2025. 31 ILO Working Paper 144 X2 Sectoral, company, and workplace social dialogue Labor unions have sought to use social dialogue at the sectoral, company, and workplace levels to influence how companies use AI and algorithm-based technologies, as well as the impacts of these technologies on workers and working conditions. In this section, we organize our discussion around our three themes – first, social dialogue over the employment and skill impacts of AI (from labor replacing to complementing); second, social dialogue over algorithmic management (from labor controlling to empowering); and third, social dialogue over work location, employment status, and value chain monitoring (from labor displacing to embedding). Some collective agreements cross-cut these three categories, and so are discussed in multiple sections. 2.1. Social dialogue over employment and skill impacts of AI: from labor replacing to labor complementing Labor unions around the world are responding to the potential threat to jobs associated with AI-based automation with job and location security agreements, restrictions on permitted uses of AI, and rules concerning ownership of and control over creative work and images. They also support institutions that promote skill upgrading and adaptation, and that enhance worker control over how they apply their skills when using AI-based tools in their jobs. These efforts seek to encourage employers to use AI in ways that complement and upskill rather than replace and deskill work. The backdrop to these efforts is the potential for algorithm and AI-based tools to support significant changes in work organization, skills, and productivity. In manufacturing, AI-enabled robots and smart machines associated with ‘Industry 4.0’ enable new efficiency and productivity gains on the shop floor. In services, robotic process automation, speech analytics, and generative AI-enabled digital assistants permit automation of an increasingly broad range of backand front-office tasks, from call centers to HR, marketing, accounting, and even coding or advanced diagnostic work. The most controversial impacts are being seen in creative industries and occupations, as generative AI tools from ChatGPT to Dall-E are being applied to automate aspects of creative writing, translation, visual art, modeling, and acting. A large body of research, much of it in economics, has established the potential employment risks associated with the accelerating adoption of AI. Originally, the term ‘task-biased technological change’ referred to the potential for automating tasks lacking creative and social intelligence, as well as tasks that require manual manipulation–with researchers debating the potential number of jobs at risk of automation.126 However, following recent advances in generative AI, nonroutine tasks in high skilled jobs are also seen to be at risk of automation, from professional services such as accounting, law, health care, and translation to journalism and art.127 While this suggests a grim future characterized by mass technologically-induced unemployment, the ways in which firms use these tools is a matter of strategic choice. AI can also be used in ways 126 Frey, C. B., & Osborne, M. A. (2017). The future of employment: How susceptible are jobs to computerisation? Technological forecasting and social change, 114, 254-280. 127 Susskind, D. (2020). A world without work: Technology, automation and how we should respond. Penguin UK. Eloundou, T., Manning, S., Mishkin, P., & Rock, D. (2023). GPTs are GPTs: An early look at the labor market impact potential of large language models. arXiv preprint arXiv:2303.10130. 32 ILO Working Paper 144 that ‘augment’ labor and complement skills - for example, through emphasizing worker choice in using tools to improve the quality of services, or restructuring production in a way that creates new, high-productivity tasks.128 Economists Acemoglu and Restrepo observed in 2019 that corporate deployment of AI in the US was biased towards automation applications focused on short-term cost savings, generating risks of increasing unemployment and inequality as well as stagnating productivity.129 They also encouraged organizations and policymakers to take a different path - to capture the promise of the ‘right’ kind of AI associated with improved economic and social outcomes. A key question these trends raise is how social dialogue can best be deployed to encourage firms to adopt AI-based tools in a way that complements human creativity and innovation–or to pursue longer-term, productivity enhancing strategies that produce mutual gains for workers, firms, and societies rather than driving up unemployment.130 An OECD survey examined the role of social partners and social dialogue in interacting with artificial intelligence. 131 It assessed social partners’ awareness of what AI entails, their assessments of the risks and benefits relating to AI adoption in labour markets and workplaces, and their responses to AI adoption. Results showed that employer associations were mostly interested in skill enhancement and productivity gains, while labor unions focused on high job quality and trustworthy usage. This suggests that social dialogue can play an important role in bridging these goals, which together can be the basis for more broadly shared social gains from AI investments. Our case study findings indicate that successful efforts by labor unions in the areas of employment and skills often build on existing institutional protections through laws, policies, and collective agreements. We also see creative new agreements and organizing efforts that mobilize workers to win AI-specific rules. These seek to strengthen constraints on or disincentives for employers to use new technologies to ‘exit’ employment relationships with their workforce through downsizing, outsourcing, or deskilling jobs. They also often draw on, while also further developing support for, collective worker voice in how AIand algorithm-based tools are used in different jobs and professions. 2.1.1. Europe In Nordic countries, many union-led social dialogue initiatives on employment and skills aspects of technological change have focused on the sectoral level. This may be attributed to their traditions of voluntarist, multi-level bargaining with local bargaining power supported through high union density and bargaining coverage. The Nordic countries are also ‘some of the most digitalized societies in the world’, and their companies have often been early adopters of AI technologies.132 In general, Nordic labor unions have focused on strengthening investments in skills while ensuring worker participation in decision-making related to employment adjustments. In Sweden, joint initiatives and agreements across sectors have often focused on ensuring the expansion of 128 Zysman, J., & Nitzberg, M. (2024). Generative AI and the Future of Work: Augmentation or Automation?. Available at SSRN 4811728. Johnson, S., & Acemoglu, D. (2023). Power and progress: Our thousand-year struggle over technology and prosperity. Hachette UK. 129 Acemoglu, D. & P. Restrepo (2019) ‘The Wrong Kind of AI? Artificial Intelligence and the Future of Labor Demand’ NBER Working Paper 25682, March 2019, https://www.nber.org/papers/w25682. 130 McElheran, K., Li, J. F., Brynjolfsson, E., Kroff, Z., Dinlersoz, E., Foster, L., & Zolas, N. (2024). AI adoption in America: Who, what, and where. Journal of Economics & Management Strategy, 33(2), 375-415. 131 OECD (2023) OECD Employment Outlook 2023. Chapter 7. https://www.oecd.org/en/publications/oecd-employment-outlook2023_08785bba-en.html 132 Ilsøe, A., Larsen, T. P., Mathieu, C., & Rolandsson, B. (2024). Negotiating about algorithms: Social partner responses to AI in Denmark and Sweden. ILR Review, 77(5), 856-868. 33 ILO Working Paper 144 high quality jobs to replace those lost through automation through industrial policy, as well as supporting life-long learning opportunities to prepare workers for changing skill demands.133 In Finland, a collective agreement for salaried ICT employees (2023-2025) established a working group to examine AI’s implications for employment conditions and worker roles. In Norway, a 2017 collective agreement in the government sector included a commitment to social dialogue and worker participation to digitalize public services in a more inclusive way. Specific reference was made to joint committees focused on digitalization or digital agents.134 And in Denmark, the 2023 agreements covering creative industries included a commitment by unions and the employer association to discuss the potential impacts of AI, and to assess adjustments needed in response.135 Germany has the largest number of, and best documented or studied, collective agreements that address AI’s employment and skill impacts. German labor unions and works councils are supported in these efforts by strong job security rights at national level and in collective agreements; as well as by recent revisions to the Works Constitution Act (discussed in Section 2.1) that extended works councils’ information, consultation, and co-determination rights on new technologies and their employment impacts to plans to adopt AI. A number of early agreements on digitalization resulted from a project dubbed ‘Arbeit 2020’, organized jointly by the Industrial Union of Metalworkers (IG Metall), the Mining, Chemicals, and Energy Industries Union (IGBEC), and the Food, Beverages, and Catering Union (NGG).136 They first studied the change process associated with Industry 4.0 in 28 selected plants, together with outside consultants, with the goal of understanding digital change and concluding agreements on how to jointly influence that change. This project resulted in 13 initial agreements, all of which included provisions strengthening plant-level skills development. One prominent manufacturing example is a framework agreement for implementation of Industry 4.0 projects at Airbus, which established a process for works councils to participate in strategy-building, rules for introducing new technologies, and guidelines for training workers engaging with those technologies.137 Managers produce project profiles describing the process and impacts of new technology introduction on employment and work processes, and a joint management-works council steering committee jointly decides on the process for technology introduction. Several agreements in service firms also address employment and skills impacts of new technologies. An agreement from 2018 between Deutsche Bahn and the Railway and Transport Union (EVG) establishes a process for worker participation in decisions associated with the planning, development, or introduction of new digital tools.138 The agreement includes provisions committing the employer to conduct an assessment of how digital innovation will affect employment and worker protection.139 133 Email communication, Victor Bernhardtz, Unionen, January 21, 2025. 134 Voss and Bertossa (2022), Ibid, p.7. 135 Ilsøe et al. (2024) Ibid. p. 861. 136 Bosch, G. and J. Schmitz-Kießler (2020). Shaping Industry 4.0–an experimental approach developed by German trade unions. Transfer: European review of labour and research, 26(2), 189-206. 137 Harbecke, T. and G. Mühge (2020) Digitalisierungsstrategien im Portrait. 34, Mitbestimmungspraxis. Düsseldorf: Hans Böckler Foundation. Krzywdzinski, Martin, Detlef Gerst, and Florian Butollo. (2023) Promoting human-centred AI in the workplace. Trade unions and their strategies for regulating the use of AI in Germany. Transfer: European Review of Labour and Research 29(1): 53-70. 138 Voss, E., & Bertossa, D. (2022). Collective Bargaining and Digitalization: A Global Survey of Union Use of Collective Bargaining to Increase Worker Control over Digitalization. New England Journal of Public Policy, 34(1), 10. p.8. 139 Tarifvertrag zur Zukunft der Arbeit im Rahmen der Digitalisierung im DB-Konzern. https://www.evg-online.org/fileadmin/Tarif/ Tarifvertraege/Tarifvertraege_DB_Konzern/TV_Arbeit_4.0_EVG_2018_internet.pdf 34 ILO Working Paper 144 At the clothing retailer H&M Germany, the United Services Union (ver.di) negotiated an agreement, which addresses declining worker commissions due to online sales and potential downsizing or deskilling due to the introduction of new digital tools in stores.140 Store workers receive new bonuses, training investments, and dismissal or demotion protections–including a guarantee that temporary workers will not replace permanent employees. The agreement also establishes a digitalization advisory board made up of union and management representatives, which collects worker feedback, advances proposals relating to technology impacts on work, and extends works council participation rights on digitalization. In addition, in February 2025, ver.di and the German actors’ union (BFFS) negotiated an agreement with the German film producers’ association (Produktionsallianz) on the use of generative AI (GAI) in film production. It includes provisions requiring actor consent for the use of digital replicas, transparency concerning how GAI is used on an actor’s performance, fair compensation for AI-generated scenes, and restrictions limiting the use of AI-generated replicas to the project they were created for.141 This agreement will be used as a starting point for negotiations concerning GAI impacts on film crews, with a joint study on this impact currently commissioned.142 Deutsche Telekom has among the most comprehensive agreements establishing job security and upskilling connected to digitalization and AI (See also case study 2).143 A 2010 agreement states that automation should first be used to reduce subcontracting, while committing the employer to train internal employees affected by automation for new jobs. In the mid-2010s, the works council organized an 8-month project to analyze the impact of new digital, algorithm, and AI-enabled tools on worker skills, jobs, and performance. Based on findings, a series of agreements was negotiated, establishing a process for ongoing consultation and negotiation over new technologies and their workforce impacts. Management committed to drawing up a ‘digi-road map’ laying out digitalization measures planned for the next few years, and then discussing with the works council the impacts of these measures on employment numbers, service quality, and work content. This feeds into strategic planning on new agreements for specific technologies. Deutsche Telekom’s agreements gave workers a baseline of job security to encourage joint labor-management efforts on AI adoption and deployment, which focused on improving service quality and productivity through investments in worker skills.144 A recent provisional works agreement at the insurance provider Provinzial is distinctive in including specific guidelines on the use of generative AI, specifically Chat GPT, for service employees. 145 These include requirements that the tool should be used to facilitate work, not to ‘dismantle employment’, that machine-generated content should be marked as ‘generated by Provinzial Chat GPT’ until it has been checked and changed by a responsible human, and that only the users of the AI tool can access their own chat histories. The agreement also states that the model test ‘is also intended to gain experience with how the use of generative AI systems affects training needs, employment, and the value and skill level (or deskilling) from the perspective of users and those affected.’ The agreement was the basis of a partnership between the works council 140 UNI Global Union. (2022) H&M Workers Protected Under First Digitalization Agreement with ver.di. https://uniglobalunion.org/news/ hm-workers-protected-under-first-digitalization-agreement-with-ver-di/ 141 Ver.di (2025) Erster Tarifabschluss zum Umgang mit KI in der Filmund Fernsehproduction. February 10, 2025. https://filmunion.verdi. de/und-action/nachrichten/++co++0a0e1f88-e78f-11ef-81ee-f5ae95fe3e8f 142 UNI Global Union (2025) First collective agreement on the use of AI in film and TV production in Germany, March 14, 2025. https:// uniglobalunion.org/news/first-collective-agreement-on-the-use-of-ai-in-film-and-tv-production-in-germany/ 143 Doellgast, V., and T. Kämpf (2023) Co-determination meets the digital economy: works councils in the German ICT services industry. Entreprises et histoire 4: 32-43. Haipeter, T., M. Wannöffel, J-T Daus, and S. Schaffarczik (2024) Human-centered AI through employee participation. Frontiers in Artificial Intelligence. 7 144 Multiple interviews, works councilor and management representatives, 2021-2024. 145 Works Agreement ‘Provinzial GPT Chat’. 35 ILO Working Paper 144 and management. The goals of this partnership included developing use cases where AI could improve productivity and quality, and sharing best practices more broadly across the organization; but under the control of the Provinzial workers. As with Deutsche Telekom, an important starting point was a job security agreement.146 Although we were able to find the largest number of examples of companyand workplace-level agreements focusing on employment and skills aspects of AI and digitalization in Germany, these do not represent a universal practice. IG Metall conducted a survey in 2019 showing that a minority of works councils in the German metal sector were involved in digitalization projects at an early stage.147 A more recent 2023 survey of 385 managers and 224 works councilors found widespread adoption of AI tools among surveyed companies, with the main goals to automate work and increase efficiency.148 Survey results suggested divided labor relations experiences, with half of surveyed companies and works councils reporting cooperative relationships on AI, and the other half describing conflictual co-determination. Consistent with the findings from the case studies discussed above, more cooperative co-determination on AI was associated with greater employee involvement in AI adoption and higher acceptance of AI technologies among the workforce. Other European countries have weaker participation rights on technologies but also overall strong institutional protections for job security, which can be a tool in social dialogue over technology-related restructuring. In France, the introduction of AI has often involved conflict, due in part to the lack of clarity concerning the role and rights of unions and works councils to consult on these technologies. In one example, the French public employment service (France Travail– formerly Pôle Emploi), sought to introduce an AI-based tool to answer emails more quickly.149 Chagny and Blanc describe how only one out of seven unions voted on the initiative, but it also demanded a study of the impact on employment, working conditions, and service. The union also called for an ethical charter, which was eventually adopted three years after management announced the AI project. Other French employers have addressed concerns with AI use through similar ‘ethical charters’. The telecom provider Orange France established a Data and Ethics Council, composed of academics and chaired by Orange’s Chief Technology Officer, which drew up an Ethical Charter in 2022 establishing principles that include respecting human autonomy, operating under human supervision, and respecting equality, diversity, and privacy.150 In 2024, the Le Monde group adopted an ethical charter on the use of AI by its newsrooms, which states that AI cannot replace humans in journalistic production, generative AI can only be used to assist editorial production ‘under strictly defined conditions’, and generative AI use to create images is prohibited.151 However, both of these appear to be unilateral initiatives that did not directly involve labor unions. Recent developments suggest the potential for strengthened social dialogue over AI at company level in France. A 2022 court decision held that ‘companies should accept that workers’ 146 Interview, Provinzial Works Councilor, October 2, 2024 147 Gerst, D. (2020). Geschäftsmodelle mitentwickeln– ein neues Handlungsfeld der Betriebsräte. WSI-Mitteilungen 73, 295–299. doi: 10.5771/0342-300X-2020-4-295 148 Krzywdzinski, M. (2024) Zwei Welten der KI in der Arbeitswelt Wie Management und Betriebsräte die Einführung und Nutzung von KI-Anwendungen gestalten. Weizenbaum Discussion Paper #39, June 2024. https://www.weizenbaum-library.de/server/api/core/ bitstreams/dd69f1ed-98ce-4b64-a996-d44521919ba3/content 149 Chagny and Blanc (2024), ibid, p.198 150 Orange (2022) Orange adopts a Data and Artificial Intelligence Ethical Charter. November 15, 2022. https://newsroom.orange.com/ orange-adopts-a-data-and-artificial-intelligence-code-of-ethics/ 151 van Kote, G. (2024) Le Monde adopts a charter on artificial intelligence. Le Monde, March 13, 2024.https://www.lemonde.fr/en/aboutus/article/2024/03/13/le-monde-adopts-a-charter-on-artificial-intelligence_6615286_115.html 36 ILO Working Paper 144 representatives (the social and economic committee [i.e. works council]) must be consulted and have recourse to an expert when new technology (in this case an AI system) is introduced, even if it has no identified impact on working conditions’ based on existing labor code provisions.152 Several sectoral agreements in other European countries focus on skill and employment topics. In Belgium, social partners in the commerce sector agreed on a ‘Memorandum for a Sustainable and Competitive Retail Sector’ in 2024.153 The document stresses the importance of social dialogue to support e-commerce and work organization concerns, and includes seven ‘strategic axes’ for policy - including training in digital skills and guaranteeing a level playing field. The Italian Banking Association (ABI) collective agreement created a joint committee to monitor the impacts of digitalization, focusing on work reorganization and new roles. Finally, worker representatives are beginning to address employment concerns at international level connected with digitalization. The Spanish multinational retail firm Inditex – famous for its Zara clothing brand – agreed a ‘Strategic Digital Transformation Plan’ with its European works council154 in 2020.155 This included commitments to ‘maintain a stable staff’ during restructuring through digitalization and online integration, with a process launched with labor unions to relocate and retrain employees. 2.1.2. North America European labor unions have mobilized their comparatively strong bargaining and job security rights to encourage employers to commit to investments in training and to limits on top-down employment restructuring associated with AI technologies. However, United States unions have been on the forefront of negotiating specific provisions in collective agreements that address the employment and skills aspects of generative AI-based tools. This is due to the leadership of labor unions in the entertainment and creative sectors. A common theme across these contracts is the principle of creative worker control over how AI is used and rules safeguarding employment to prevent narrow ‘labor replacing’ uses of these tools. While the US does not have the same tradition of sectoral collective agreements as in many of the European case study examples, there continues to be sectoral bargaining in the ‘gig based’ entertainment industries – including for screenwriters, actors, and musicians. The Writers Guild of America (WGA) was a trailblazer in the area of AI agreements. Following a 148 day strike, in which the potential threats of AI to screenwriters’ job security and control figured prominently, the union negotiated specific restrictions on the use of AI for script writing in the 2023 Theatrical and Television Basic Agreement. These included provisions stating that AI cannot be used to write or rewrite literary material, AI-generated material cannot be considered 'source material', writers cannot be required to use AI, and AI-generated material must be 152 Chagny and Blanc (2024), ibid, p.199 – citing TJ Pontoise, 15 April 2022, no. 22/00134. 153 Mémorandum pour un secteur de la distribution durable et compétitif. https://www.ccecrb.fgov.be/p/fr/1208/memorandum-pourun-secteur-de-la-distribution-durable-et-competitif/11 154 European works councils (EWCs) are bodies that represent the European employees of companies with at least 1000 employees in the EU and the European Economic Area, and with at least 150 employees in each of two Member States. The process of creating an EWC is triggered through a request by 100 employees from two countries or an employer initiative. EWCs have information and consultation rights at transnational level. See: European Commission (2025) European Works Councils. https://employment-socialaffairs.ec.europa.eu/policies-and-activities/rights-work/labour-law/employee-involvement/european-works-councils_en 155 Inditex (2020) Joint statement by the Inditex group and the European Works Council about the Digital Transformation Strategic Plan https://www.uni-europa.org/old-uploads/2020/12/EN_-DECLARACION-CONJUNTA-INDITEX-Y-CEE-ANTE-EL-PLAN-ESTRAT%C3%89GICODE-TRANSFORMACI%C3%93N-DIGITAL.pdf 37 ILO Working Paper 144 disclosed. Additional provisions asserted the union’s role in enforcing copyright and legal prohibitions on using writers’ material to train AI.156 The 2023 SAG-AFTRA collective agreement similarly addressed the use of generative AI – in this case to produce digital replicas of a performer’s voice or likeness. Case study 1 details the labor conflict leading up to this agreement and outcomes concerning AI regulation, as well as ongoing conflicts involving SAG-AFTRA represented actors in the game development industry. XCase study 1: SAG-AFTRA – US actors mobilize to establish AI guidelines in film, television, and game development In 2023, the Screen Actors Guild-American Federation of Television and Radio Artists (SAGAFTRA) went on strike for 118 days, resulting in a collective agreement with the Alliance of Motion Picture and Television Producers (AMPTP) that included first-of-its-kind protections against the use of AI-generated replicas of human performers.157 Provisions were based on two fundamental principles. First, digital replication of a performer’s voice and/or likeness was prohibited without advance notice and written informed consent. Second, in order to reduce economic incentives to replace human performers with replicas, it required that performers be paid the same amount for the use of their digital replica as they would be for a human performance. At the time of the agreement, generative AI was not sufficiently advanced to fully replace human performers, and instead was mainly being used to de-age actors, modify lip movement for more realistic dubbing, and produce audio books. Voice actors are especially vulnerable to generative AI because voice cloning technology is already 'good enough' at believably simulating a human voice. Knowledge of AI proved to be a key source of leverage for SAG-AFTRA. The union met with around thirty AI companies specializing in digital replication, including OpenAI, to see demonstrations of the technology, to find out how easily and how well the technology could simulate human performance, and to get a sense of where the industry was heading in the near future. An AI working group formed in early 2023, made up of staff with expertise in voice, likeness, and intellectual property law. The group continued to meet regularly to share stories and discuss contract language, articles, and case studies. An AI task force was also formed, where rank-and-file members with above-average knowledge of technology could interact with union staff and discuss AI-related issues in their workplaces. SAG-AFTRA’s extensive and proactive information gathering have supported its campaigning, bargaining, and representation activities; as well as union representatives’ capacity to advise the studios, which now view the union as a resource in their AI investment and deployment decisions. Policymakers are also drawing on SAG-AFTRA’s AI expertise. The union was instrumental in drafting the No Fakes Act, which would guarantee a federal right to voice and likeness; and ensuring it was introduced in both the House and the Senate. Studios, record labels, broadcasters, tech companies, artists, advocacy groups, and the AFL-CIO supported the bill, which also has bipartisan support. SAG-AFTRA is also directly involved in similar legislative 156 Writers Guild of America (2023) Summary of the 2023 WGA MBA. 157 Interview with SAG-AFTRA Executive VP and General Counsel, November 6, 2024. 44 ILO Working Paper 144 a way as to mandate direct employment for call center workers in the financial sector, ensuring better protection of customers' financial information. In Brazil, the banking sector has also been the focus of worker mobilization and collective bargaining on AI, but with an emphasis on investments in equity in training and job opportunities. The proportion of women working in the Brazilian banking sector was declining due to the automation of customer service work and the shift to online banking, while new jobs were expanding in more technology-intensive IT jobs. CUT’s National Confederation of Financial Workers (CONTRAF-CUT) thus began an initiative to expand training to women to become specialists in this area, to encourage them to move into more specialized positions.186 This was organized in partnership with training colleges and banking employers, with the goal of making scholarships to IT courses available to women. In 2024, CONTRAF-CUT signed a national collective agreement with the National Confederation of Banks (FENABAN), introducing clauses on AI and workers’ training.187 The agreement includes commitments to support retraining and promotion of opportunities ‘in the face of technologies such as AI’, providing information to workers on initiatives and disruptive technologies (including AI), and mitigating unequal pay – with a focus on gender equality. It establishes a monitoring process for retraining initiatives, which the agreement states will be carried out through the ‘National Negotiation on New Technologies, such as AI, and Banking’.188 The banking employers also agreed to finance 3,000 scholarships for women to take IT courses focusing on data analysis, programming languages, and web design. Priority will be given to women ‘in situations of socio-economic vulnerability’, and inclusive of trans women, black women, and people with disabilities.189 In South Africa, again banks have been the focus of union organizing on AI-related restructuring. The Congress of South African Trade Unions (COSATU)-affiliated South African Society of Bank Officials (SASBO) has argued that AI is potentially disruptive to jobs and careers.190 SASBO has challenged restructuring in the South African Banking Sector, arguing that employers have used AI to justify unnecessary branch closures.191 In September 2019, SASBO called a nationwide strike to protest mass retrenchments by the ABSA group, Nedbank and the closure of 91 branches by Standard Bank. The union demanded that the banks come to the negotiating table, consult the union on workforce changes, and redeploy laid off employees into new roles.192 The strike was enjoined by the labor court. Negotiations following the strike led to a voluntary severance package for employees, a salary increase and redeployment of employees who did not wish to leave their jobs to new positions by December 2019. However, union officials were unable to bargain directly over the introduction of new technology.193 186 Presentation, Juvandia Moreira, President CONTRAF-CUT, U.S.-Brazil Commercial Dialogue: Digitalization & Workforce Development Webinar. August 21, 2024. 187 Convenção coletiva de trabalho - FENABAN & CONTRAF https://contrafcut.lumis.com.br/data/files/7F/A5/5D/44/ BB102910A7809E19820808A8/Fenaban%20-%20CCT%20Geral%202024-2026.pdf 188 Translation and information on the agreement was provided by Jonas Valente, Fairwork Project, University of Oxford - email communication, February 24, 2025. 189 SindBancarios (2025) Mais mulheres na TI: abertas inscrições para cursos reivindicados pelo movimento sindical, March 12, 2025. https:// www.sindbancariospetropolis.com.br/noticia/mais-mulheres-na-ti-abertas-inscricoes-para-cursos-reivindicados-pelo-movimentosindical/12103 190 Bischoff, C., Kamoche, K., & Wood, G. (2024). The formal and informal regulation of labor in AI: The experience of Eastern and Southern Africa. ILR Review, 77(5), 825–835. 191 Kulkarni, P. (2019). South African labor court interdicts country-wide strike by bank officials. People’s Dispatch. https://peoplesdispatch. org/2019/09/27/south-african-labor-court-interdicts-country-wide-strikeby-bank-officials/ 192 SASBO The Finance Union (2019). Sasbo is challenging the banks on 4IR issues. SASBO News. November, 2019. https://www.sasbo. org.za/wp-content/uploads/2020/04/SASBO-News-v41n4-Proof-27-Nov.pdf 193 Maake, S. (2022). Exploring SASBO’S response to the challenges of artificial intelligence in the banking industry in Gauteng (pp. 70–74) [Thesis, University of Johannesburg; Master of Arts (MA)]. https://hdl.handle.net/10210/499894 45 ILO Working Paper 144 2.2. Social dialogue over algorithmic management: From labor controlling to labor empowering Management itself is being transformed by tools that automate scheduling, work allocation, monitoring, and performance management. The term ‘algorithmic management’ describes the use of software algorithms to enable different forms of automated or semi-automated management decision-making.194 AI is increasingly integrated into coaching software, and is used in predictive or ‘human resource’ analytics to hire new workers, determine training needs, and allocate work. More conventionally, AI is used to recognize patterns recorded or gathered via diverse electronic data sources to evaluate performance. This includes, for example, AI-enabled cameras applying machine vision, which allows for automating employee monitoring through ‘anomaly detection’; wearable devices that apply AI algorithms to analyze patterns in biometric information, employee movements, and location; and voice monitoring technologies that apply AI-based speech analytics and sentiment analysis to identify adherence with scripts, evaluate customer service quality, and direct or coach employees during service interactions. There are distinct risks to workers associated with algorithmic management, including privacy of personal data, bias, and discrimination, but also work intensification, increased surveillance, and reduced discretion or control at work. One concern is that these tools gather and use employee data. Thus, data privacy may be at risk, particularly if there are no or weak guidelines concerning which kinds of employee data the software can access and how it will be used.195 Second, these technologies can encourage intensification of management control. Tools like speech analytics or computer vision require constant monitoring of employee performance to recognize patterns and detect ‘anomalies’. Biometric or GPS-based tools used in warehouses, retail, or housekeeping can tightly control worker movements and remove their discretion over the order and manner in which they perform tasks.196 AI-based coaching technologies in customer service settings often are adopted in a context in which employees are required to follow tight scripts or to receive constant feedback about their performance.197 Third, these tools can remove certain decisions from humans, so workers have limited ability to complain or contest unfair decisions.198 Together, these add up to substantial risks to worker health and well-being. Past research has shown that management practices based on intensive monitoring, restricted control over work, and perceived lack of transparency and fairness in work allocation or performance evaluation decisions are connected to worker stress and burnout.199 Labor unions have sought to reduce these risks in social dialogue with employers through securing stronger rights for collective worker voice over how algorithmic management tools are used at work. These efforts often draw on existing data protection and participation rights, or focus on establishing them in collective agreements. Unions have also sought to establish more equitable policies concerning workplace applications of algorithmic management tools, through campaigns organized in solidarity with 194 Wood, AJ (2021) Algorithmic management consequences for work organisation and working conditions. JRC Working Papers Series on Labour, Education and Technology, No. 2021/07. 195 Ajunwa, I. (2023). The quantified worker: Law and technology in the modern workplace. Cambridge University Press. Aloisi, A., & De Stefano, V. (2022). Your Boss Is an Algorithm: Artificial Intelligence, Platform Work and Labour. Bloomsbury Publishing. 196 Delfanti, A. (2021). The Warehouse. Workers and Robots at Amazon. Pluto Books. Kassem, S. (2023). Work and Alienation in the Platform Economy: Amazon and the Power of Organization. Policy Press. Vallas, S.P., Johnston, H. & Mommadova, Y. (2022) Prime suspect: mechanisms of labor control at Amazon's warehouses. Work and Occupations, 49(4), 421–456. 197 Doellgast, V. (2023). Strengthening social regulation in the digital economy: comparative findings from the ICT industry. Labour and Industry, 33(1), 22-38. 198 Dupuis, M. (2025). Algorithmic management and control at work in a manufacturing sector: Workplace regime, union power and shopfloor conflict over digitalisation. New Technology, Work and Employment. 40(1): 81-101 199 O’Brady, S. and Doellgast, V. (2021) Collective Voice and Worker Well‐being: Union Influence on Performance Monitoring and Emotional Exhaustion in Call Centers. Industrial Relations: A Journal of Economy and Society 60(3): 307-337. Pfeffer, J (2018) Dying for a paycheck: How modern management harms employee health and company performance —and what we can do about it. HarperBusiness. 46 ILO Working Paper 144 those workers most negatively impacted by them. We outline examples of these efforts in this section, with a focus on ‘traditional’ workplaces. We do not discuss examples of agreements or social dialogue over algorithmic management in platform and app-based work, as these are covered in section 2.3 below. 2.2.1. Europe Europe has led the way on agreements focusing on algorithmic management, with many negotiated provisions limiting the use of AIor algorithm-based tools for worker surveillance and performance management. At EU level, as discussed above, the General Data Protection Regulations (GDPR) has been in force since 2018, with requirements concerning the collection, storage, and use of personal data. More recently, the AI Act establishes some further protections, including prohibitions of automated decisions that affect employees without human oversight; and on the use of AI for sentiment analysis or ‘emotion recognition’ – effectively banning a controversial AI application that is increasingly used in AI-based hiring and coaching tools. Nordic unions have negotiated most publicly over broader training or partnership initiatives related to AI at industry or national level, as discussed in previous sections. However, several case studies show company-level bargaining over specific issues around algorithmic management. In Norway, a major telecommunications provider negotiated a series of agreements with its unions restricting the use of video monitoring and speech analytics, following a conflict over increased digital surveillance in the company’s call centers.200 The union escalated its complaints over video monitoring to the Norwegian Data Protection Agency, which found that this violated employee privacy rights. After this, the unions and management reestablished more cooperative negotiations, including consultation on the use of a training platform applying real-time analytics to identify skill gaps, and participation in a taskforce to consult on an AI-based speech analytics tool, which restricted its use for identifying individual performance data. In Sweden, unions’ strategies have focused on applying existing regulations and protections to local negotiations on AI-based issues. A Unionen representative observed, for example, that occupational health and safety regulations include rules concerning negotiations over new work tools and risk assessment that can be used to support worker voice in algorithmic management tools.201 Similarly, the Co-Determination Act requires employers to negotiate where there is any significant change to an operation or to employees’ working or employment conditions. As negotiations must be called well in advance of changes being implemented, trade unions could use their rights under this legislation to influence the implementation of AI systems ex ante.202 The union’s strategy (and challenge) was thus to strengthen local capacity to use these rights in negotiations. A case study from Sweden documents an example of this local involvement, or what Bender and Söderqvist describe as ‘technological co-determination’, in the mining and smelting company Boliden.203 Here negotiations between the employer and its unions focused on a wifi positioning system. Local collective agreements formalized an agreement worked out through consultation: workers were to be given anonymous identification numbers that could only be de-anonymized 200 Doellgast, V., Wagner, I., & O’Brady, S. (2023). Negotiating limits on algorithmic management in digitalised services: cases from Germany and Norway. Transfer: European Review of Labour and Research, 29(1), 105-120. 201 Email correspondence with Victor Bernhardtz, Unionen, November 14, 2024 202 Email correspondence with Victor Bernhardtz, Unionen, January 21, 2025 203 Bender, G. and F. Söderqvist (2024) Human-Centered or Biorobotized Automation? Technological Codetermination in an Innovative Mining Company. Draft Paper. 47 ILO Working Paper 144 under emergency procedures – in which case a union representative was required to remove the anonymization feature using a password. Germany again is the exemplary case, with numerous agreements regulating the use of AIor algorithm-based tools for performance management, the direction of tasks, and workforce analytics. In Case study 2 below, we discuss in more detail two agreements negotiated in the German ICTS industry, which have strengthened worker voice in how algorithmic management tools are used in service workplaces. A key focus in both has been establishing AI Ethics Committees to oversee the ethical adoption of AI tools, as well as to establish and enforce clear rules prohibiting certain uses of AI that pose significant risks to workers and working conditions. XCase study 2: Negotiating ‘labor empowering’ agreements in the German ICTS industry: Deutsche Telekom and IBM204 Germany has very strong worker rights both to data protection and participation in decisions concerning technology at work, as discussed above.205 In addition, the federal Works Constitution Act gives works councils strong co-determination rights over the use of technologies for ‘performance and behavior’ control. These rights have been a key tool in negotiations over algorithmic management. In the information and communications technology services (ICTS) industry, German works councils – supported by the labor unions ver.di and IG Metall – have negotiated agreements strengthening worker voice over how technologies are used in algorithmic management, particularly for monitoring worker performance but also for a range of algorithmic management applications. Deutsche Telekom in Germany negotiates collective agreements with ver.di and works agreements with its elected works councils. Together, these have established a series of rules and provisions (discussed above) that encourage ‘labor complementing’ use of AI, through committing management to job security and retraining as jobs are cut due to automation; and committing management to drawing a roadmap of digitalization measures with employment impacts. In addition, a central focus of negotiations has been to support worker privacy, mitigate bias, and discourage the use of AI tools that intensify surveillance and top-down management control of workers. A works agreement on IT systems establishes a process through which management consults with the works council before purchasing new technology. Based on an evaluation of the risk to employees, a joint labor-management committee decides whether formal negotiations are needed. An important criterion for evaluating ‘risk’ is whether the software can record or track data on individual employees – as all performance metrics must be aggregated to groups of five employees or more. An agreement on workforce analytics states that analytics tools should be used to improve the working environment and support management decisions. It prohibits using analytics to monitor individual employee performance or behavior or to make automated decisions 204 Some of the text in the summary of the Deutsche Telekom case is taken from pp.48-49 the report: Doellgast, V., O'Brady, S., Kim, J., & Walters, D. (2023). AI in contact centers: Artificial intelligence and algorithmic management in frontline service workplaces. Cornell University. https://hdl.handle.net/1813/113706 205 Germany has had a federal data protection act since 1978. The current Federal Data Protection Act (BDSG) was revised to conform with the EU’s General Data Protection Regulation (GDPR). 48 ILO Working Paper 144 without human oversight. Databases must be designed so that all data are anonymous, and no conclusions can be drawn about individual employees. Transparency under data protection laws must be ensured, particularly regarding mathematical-statistical (or algorithmic) processes. Special measures are required for certain categories of personal data, with reference to the GDPR. The agreement establishes a ‘Workforce Analytics Expert Group’ with equal works council and employer representation. It reviews the use of employee data and AI-enabled analytics tools, holding regular evaluation workshops and including provisions for training employees (with works council involvement) to use workforce analytics responsibly. The works council also drafted an AI ethics manifesto stipulating how AI would be used and recommending that an expert group be established to ensure adherence to the manifesto’s principles. It has several key provisions: Interactions with ‘learning machines’ must be designed so that workers know that they are interacting with AI, and human decision-makers (not AI) must make significant personnel-related decisions. Certain uses of AI are prohibited, including gathering personal information about employees’ political opinions, philosophical beliefs, union membership, or sexual orientation; or seeking to analyze or influence employees’ emotions or mental state. Technologies that ascribe personality traits or use biosensors are not allowed. Works agreements at Deutsche Telekom have historically limited the use of remote monitoring technologies, and newer agreements prohibit the use of speech analytics to monitor employee emotions as well as the direct implementation of AI-enabled coaching apps. An important provision is that individual performance data can be gathered only for groups of at least five employees. Employees can see their own performance data but choose whether to share this with team leaders, who are prohibited from requesting it. Further workforce management tools have been regulated under the above agreements. For example, managers cannot access employees’ Outlook calendars, only information that is important for projects. WebEx videoconference tools cannot measure working time. Employees are not required to turn on their cameras for meetings, and there can be no recording of team meetings and training. Agreements regulate when team leaders listen to employee calls and for how long. IBM Germany also has negotiated both collective agreements with ver.di and works agreements with its elected works councils. Similar to Deutsche Telekom, past agreements established that employers could not carry out ‘behavior and performance reviews’ of individuals using data collected through the IT system. A works councilor observed that the integration of AI into these management systems was the starting point for a 2020 agreement on the use of AI systems.206 Similar to what occurred at Deutsche Telekom, this AI Framework Agreement was first discussed in joint labor-management workshops, which applied design thinking methods. 207 The agreement classifies different AI applications according to their risk and defines standards for evaluating the workforce impacts of AI. Thus, for example, the use of AI for personnel measures or automated decisions with immediate effects on employees, without human oversight, are classified as ‘high risk’ and prohibited. Algorithms that provide 206 Interview, IBM works councilor, November 8, 2021. 207 Gergs, H.-J., L. Schatilow, B. Langes, T. Kaempf (eds) (2023) Human Friendly Automation: Arbeit und Künstliche Intelligenz neu denken. Frankfurter Allgemeine Buch. 49 ILO Working Paper 144 information that is visible only to the employee and direct supervisor, such as training recommendations, are classified as ‘low risk’ and permitted. In the middle are applications, such as career planning, which may have risks for employees and thus require additional oversight or negotiation. The agreement also establishes an AI Ethics Council made up of AI experts and employer and employee representatives, which evaluates AI applications and oversees implementation of the agreement. The Council’s role is to further develop the framework agreement and advise the human resources department and the works councils on current and planned AI applications. It also discusses and debates any objections or complaints from employees concerning AI tools in the workplace. Finally, it reviews AI recommendations and corrects them, if necessary. Works councilors at both Deutsche Telekom and IBM felt that these agreements had improved workplace health and safety, while encouraging more positive workforce views of the AI-based systems that were adopted. Employees were protected in their existing agreements from invasive monitoring or privacy abuses associated with algorithmic management. The new agreements established a process for co-determination and oversight over the future adoption of workforce analytics and other tools based on clear principles and an expert-based review process. This also reduced the risk of works council opposition to expensive IT systems, standardizing the co-determination process and improving workforce trust in how managers were using potentially controversial tools, such as employee call recording and speech analytics. Together, these cases demonstrate the potential for bipartite negotiations, in a context of strong data protection rights and workers’ bargaining rights over technology, to encourage ‘labor empowering’ applications of algorithmic management in the workplace. Other cases in Germany have placed some similar limits on AI use in algorithmic management. A 2020 agreement between Amazon Alexa and its works council restricts the use of algorithm-generated performance data for HR decisions, ensuring this data is not used to promote or dismiss workers.208 This followed a dispute which was brought to an arbitration board. At Siemens, the works council established guidelines on data protection, data storage, and basic ethical considerations.209 It also developed an informal agreement with management to develop ‘AI cards’ outlining the uses of different AI technologies in the company. A number of anonymized case studies have examined works council involvement in negotiating or consulting over algorithmic management in different industry settings in Germany. For example, Wotschack et al. examine co-determination in a mechanical engineering company and a large food-delivery company.210 They find that works councils in both cases faced challenges in understanding, monitoring, and shaping the use of algorithmic systems. At the same time, they found strong union support on legal issues or questions related to technology - including bringing in external experts, using provisions in the Works Council Modernization Act that supports hiring an external AI expert. Another case study by Krzywdzinski et al. in a logistics company examines negotiation in different phases of algorithmic management, as the employer deployed 208 UNI Global Union (2023) Ibid, p. 14. 209 Haipeter, T., Wannöffel, M., Daus, J. T., & Schaffarczik, S. (2024). Human-centered AI through employee participation. Frontiers in Artificial Intelligence, 7. 210 Wotschack, P., Butollo, F., & Hellbach, L. (2024). Ibid. 50 ILO Working Paper 144 software automating storage, picking, control, and inspection processes in partnership with the software developer.211 They found that negotiations with the works council established agreements that prevented management from using the new tools to reduce employment and increase performance targets. However, they did not find ongoing worker involvement in implementing the system, which they explain as a result of ‘information asymmetries created by the algorithmic management system’.212 Sectoral collective agreements in Southern Europe have established worker rights regarding algorithmic management, with a focus on ensuring recourse to human decisions or clear limits on specific software applications. A 2022 agreement between Italian unions and the telecommunications companies Wind and TIM places limits on the use of Afiniti Advanced Routing, a tool that uses AI to match customers with call center agents.213 Companies are not permitted to use the software to monitor individual performance data or for worker surveillance. In Spain, a 2021 collective agreement in the banking sector between the Spanish Banking Association (AEB) and the labor unions CC.OO.-Servicios, FeSMC-UGT and the FINE Banking Federation includes a section on AI rights, which makes specific reference to algorithmic management. 214 It establishes worker rights not to be subject to decisions based solely and exclusively on automated variables; rights to non-discrimination; and the ability to request human oversight and intervention. It also requires management to inform worker representatives about the use of data analytics or AI systems when decision-making in HR and labor relations is based on digital models. Some similar rights are in the chemical sector agreement, which requires management to provide information to worker representatives on the objectives of AI systems; an analysis of their impacts on working conditions; and the parameters, rules, and instructions on which AI algorithms or systems are based that affect decision-making in ways that may impact working conditions or access to and maintenance of employment. AI system adoption should also involve updating occupational risk assessments, as well as ensuring there is no prejudice or discrimination and that the guiding principle is human control. In the UK, the Royal Mail Group (RMG) and the Communication Workers Union (CWU) negotiated a 2020 agreement titled the Key Principles Framework Agreement (The Pathway to Change), which established principles on union involvement in the introduction of new methods, technology, or automation. The agreement first recognized that technology was crucial to the company’s growth, innovation, efficiency, and manageable workload. It set up a series of principles for new technology introduction, with the Local Manager and Representative responsible for operational decision-making and respect for individual privacy rights. Consultation between the CWU and the RMG is required at the concept design stage. Trials cannot exceed 90 days and are set up with strict terms of reference covering content, location and success criteria. Once success criteria are demonstrated as met, this will trigger deployment, subject to business case approval.215 211 Krzywdzinski, M., Schneiß, D., & Sperling, A. (2024). Between control and participation: The politics of algorithmic management. New Technology, Work and Employment. 212 Krzywdzinski et al. (2024), Ibid, p.16. 213 Guaglianone, L. (2024) Collective bargaining and AI in Italy. In Artificial Intelligence, Labour and Society. A. Ponce del Castillo (Ed.). ETUI. 207-215. 214 Resolución de 17 de marzo de 2021, de la Dirección General de Trabajo, por la que se registra y publica el XXIV Convenio colectivo del sector de la banca. https://www.boe.es/diario_boe/txt.php?id=BOE-A-2021-5003 215 Communication Workers Union. (2020). Key Principles Framework Agreement. December 20, 2020. https://www.cwu.org/news/rmgcwu-key-principles-framework-agreement-the-pathway-to-change/ 51 ILO Working Paper 144 This agreement was followed by another agreement in 2021 negotiated between Parcelforce (a subsidiary company of RMG) and the CWU.216 It establishes that delivery drivers have a right to privacy and to the data collected by Telemetry tools; and that new technology will not be deployed for use as a disciplinary tool, or as a source of information to enhance the ability of managers to take disciplinary action. The focus should be on correction and improvement, not punishment. The Table of Success Transport Working Group will monitor key measures captured by the tools–for example, accelerating, braking and cornering–to ensure fuel efficiency and incident reduction. And training will be provided to all employees who are required to use the telemetry technology. An important provision holds that local CWU representatives will receive the same training as Depot Managers on monitoring and reviewing telemetry technology, thus supporting knowledge and capacity building within the union to negotiate over these tools and represent workers in the case of grievances. While other unions have not been able to achieve inroads into the review of algorithmic management systems to such a degree, there has been significant progress. A collective agreement between the GMB union and the delivery company Hermes required that the company reprogram its automated payment system to ensure that workers receive at least the minimum wage and that they are automatically paid any bonuses they have earned, rather than having to claim them retrospectively.217 2.2.2. North America In the United States, there are a number of examples of agreements that limit algorithmic management in service workplaces. Many of these are based on adapting past collective agreements developed for earlier monitoring technologies that are now being supplemented with AI. The CWA has negotiated agreements at major telecommunications and airline companies that restrict the use of performance data to discipline employees if they do not meet certain time-based measures such as adherence to schedule or average call handling time in call centers and that require the use of this data primarily for training purposes.218 This protects workers from unfair discipline as work volume and content changes – for example, due to expanded use of chatbots and agent assistance tools. The UPS Teamsters 2023 National Contract addresses some similar challenges concerning worker surveillance using AI tools for drivers; for example, the increasing use of infrared sensors to track driver eye movement.219 UPS is not permitted to install new driver-facing cameras or to record incab activities, while previously installed cameras must be disabled. In addition, GPS, telematics, and forward-facing cameras cannot be used as the sole basis for discipline. UPS also committed to providing workers with data from a new payroll system they plan to develop in 2026, so that they can have oversight over the basis for these decisions and management adjustments. A union 216 Communication Workers Union. (2021, August 25). LTB 354/21 - National Agreement between Parcelforce Worldwide and the CWUPilot deployment and use of driver behaviour technology. CWU. https://www.cwu.org/ltb/ltb-354-21-national-agreement-between-parcelforceworldwide-and-the-cwu-pilot-deployment-and-use-of-driver-behaviour-technology/ 217 Collins, P., & Atkinson, J. (2023). Worker voice and algorithmic management in post-Brexit Britain. Transfer: European Review of Labour and Research, 29(1), 37–52. https://doi.org/10.1177/10242589221143068 218 CWA (2014) CWA Contract Provisions to Promote Good Working Conditions for Customer Service Workers, https://cwa-union.org/sites/ default/files/6-cwa-issue-briefs.pdf. For background on call center workers’ experience of AI, see CWA, How AI is Impacting Customer Service Professional, 2024, https://cwa-union.org/sites/default/files/2023-11/call_centers_and_new_technology_fact_sheet.pdf drawing from V. Doellgast, S. O’Brady, J. Kim, D. Walters (2023) AI in Contact Centers: Artificial Intelligence and Algorithmic Management in Frontline Services. https://ecommons.cornell.edu/server/api/core/bitstreams/a0ac9f50-5a22-4b3d-a9d9-2cc06824e31d/content. For a detailed summary of (and links to) CWA contract provisions, see: Kresge, L. (2020) Ibid, pp. 15-18. 219 Kresge, L. (2023). Negotiating Workers’ Rights at the Frontier of Digital Workplace Technologies in 2023. UC Berkely Labor Center. 52 ILO Working Paper 144 representative observed that these provisions have lessened the amount of micromanaging and use of discipline, encouraging more focus on training and proactive coaching.220 The use of algorithmic management in hotel catering and cleaning has been particularly widespread, and associated with intensified surveillance as well as decreased worker control over how they carry out often physically intense work - contributing to work intensification.221 In Case study 3 below, we discuss in more detail campaigns by the Las Vegas Culinary Union to protect these workers’ privacy and expand their control over algorithmic management tools. XCase study 3: Regulating algorithmic management in Las Vegas casinos and hotels: the Culinary Union222 The Las Vegas Culinary Union – UNITE HERE Local 226, representing over 60,000 casino workers, was able to transform labor controlling technology into an opportunity for labor empowerment. In the Culinary Union’s 2018 contract negotiations with 34 casino resorts, technology was a top three issue that the membership was willing to strike over. As a result, the union won historic language in its collective agreement that included a mid-cycle process to bargain over the implementation of new technology. This has been used in negotiations over a variety of technologies, including algorithmic management tools. When one of the employers implemented algorithmic management software in housekeeping, the union was ready. Housekeepers had always been able to pick the order in which they cleaned their assigned rooms. A new smartphone app was then introduced that set the order for them, which many workers experienced as both limiting their own discretion over how they performed their jobs but also work intensifying, as they were required to work in often inefficient ways. Whenever housekeepers interacted with the app, they generated data about their workload, location, and room status. Thus, as part of the bargaining process, the union requested that backend data. One union member’s son used his programming knowledge in SQL and Python to write a script that scanned through the data to identify instances where a housekeeper had cleaned more rooms than the room quota specified in the collective agreement – which constituted contract violations. Within a couple of months, hundreds of instances were found, each with a corresponding date and timestamp. At the same time, housekeepers began keeping paper records of where the smartphone app sent them. These records were meticulously compiled and turned into maps that showed the inefficiency of the zig-zagged, algorithm-generated paths. Housekeepers then demanded that the app’s developers attend bargaining sessions. They showed the developers the maps they had made, explained what a union contract is, and told them, 'if your technology is going to be a contract violation machine, it’s going to be a big problem. So, let’s fix that.' Through worker mobilization and innovative data analysis, the workers and the employer agreed to collective agreement provisions that restored worker autonomy and discretion and made it easier for the union to monitor workload violations. In the words of one 220 Interview with Principal Officer and Secretary Treasurer, Teamsters Local 2, November 5, 2024. 221 Rho,H-J & C. Riordan (2024) Beyond algorithmic control: exploring self-sequencing of tasks in hotel housekeeping work. Draft paper presented at the ‘AI and the Future of Work Conference’, Cornell University, October 11-18, 2024. 222 Interview with CU Researcher, November 18, 2024. 53 ILO Working Paper 144 Culinary Union researcher, 'inside every worker is a nerd and inside every IT person is a worker. If people just let the IT nerds talk to the worker nerds, good things would happen'.223 Some workers in the hospitality sector have limited digital literacy. With the support of their parent union and a team of academics from four research universities, the Culinary Union formed a partnership with the app developer and the Culinary Academy of Las Vegas, a worker training center that is funded through the CWU collective agreement, to provide substantial training on the software as part of the Academy’s housekeeper training program.224 Direct contact between Culinary Union members and technology companies, as well as the use of information requests to access, organize, and analyze data generated by bargaining unit work, are innovative tactics that suggest the need for union-built technology that can enable contract enforcement. Whether it is notes jotted down on paper, photos taken by workers, or proprietary software, systematized documentation of the workplace is not the sole purview of employers. Unions can also provide resources for those who have experience with software design and data visualizations, whether they are staff, members, or other union supporters. In 2023, the Culinary Union built on the technology provisions from the previous contract and won rights to privacy from tracking technology, to bargain over technology that tracks the location of employees, to notification and bargaining over data sharing with a third party, to health care and severance pay for workers laid off due to new technology, and the right to compensation for tipped employees if tech failure makes it impossible to do their job.225 This case illustrates the important role that three-way social dialogue between labor unions, app developers, and employers can play in developing AI technologies that empower workers, with potential shared benefits in terms of improved worker autonomy and productivity. It also demonstrates the value to unions of negotiating rights to information on how workforce data are collected and used, and then to using that data to study the worker and organizational impacts of new technologies. In the US context of weak legislated data rights, the Culinary Union negotiated their own framework of these rights in a multi-employer agreement ─ strengthening collective worker voice in algorithmic management. Several agreements in US professional sports address employer collection and use of biometric data. In basketball, the NBA established a joint advisory committee to review the use of wearable technologies and data derived from wearable sensors.226 The 2020 agreement between the NFL and NFLPA established a 'Joint Sensors Committee' to oversee the use of sensors and biometric data in professional football, and requires player approval before biometric data can be 223 Interview with CU Researcher, November 18, 2024. 224 Guillot, A. (2024) The Culinary Academy of Las Vegas and Amadeus Join Forces to Expand Hospitality Employment and Job Training. Amadeus. https://www.amadeus-hospitality.com/insight/the-culinary-academy-of-las-vegas-and-amadeus-join-forces-to-expandhospitality-employment-and-job-training/ 225 UNITE HERE Local 226, press release, Culinary Union celebrates historic wins for workers in the best contract ever won with MGM Resorts, Caesars Entertainment, and Wynn Resorts, November 10, 2023, https://culinaryunion226.org/news/press/culinary-unioncelebrates-historic-wins-for-workers-in-the-best-contract-ever-won-with-mgm-resorts-caesars-entertainment-and-wynn-resorts. 226 NBA-NBPA Collective Bargaining Agreement (2017), pp.359-361. 3c7a0a50-8e11-11e9-875d-3d44e94ae33f-2017-NBA-NBPA-CollectiveBargaining-Agreement.pdf Cited in: Kresge, L. (2020). Union Collective Bargaining Agreement Strategies in Response to Technology. UC Berkeley Labor center Working Paper. Pp. 16-17. 60 ILO Working Paper 144 enough control over contract staff to be considered a ‘joint employer’. The NLRB has rejected this line of reasoning twice already, however Google is appealing both rulings in federal court.255 In 2021, the AWU organized support for a worker fired from a Google data center in South Carolina, who was employed by the subcontractor Modis, part of the Adecco Group. She worked in a job that involved fixing servers - alongside permanent Google employees - and was suspended for speaking out against unequal conditions. Following pressure from AWU, her suspension was overturned.256 In April 2023, forty-one Youtube Music workers hired via Cognizant Technology Solutions Corp. joined AWU-CWA in a unanimous vote. Less than a year later, the entire team was abruptly laid off in what both Google and Cognizant claimed were 'pre-planned contract expirations'. However, workers stated that their contracts had been routinely renewed for years prior and that they believed they were being retaliated against for union activities.257 Similarly, when Google Help workers contracted through Accenture launched their own unionization campaign, they were quickly met with waves of layoffs. Some workers were forced to train their overseas replacements before their jobs were cut. Staff had been reduced from 120 to less than 30 since the beginning of the unionization efforts in June 2023. 258 Nevertheless, AWU secured its first, historic, contract with Accenture in December 2024, which included provisions guaranteeing fully remote work, 30-days’ notice and six weeks of severance pay for layoffs, just-cause protections, and a prohibition on keystroke or mouse monitoring software.259 AWU’s Executive Board Secretary stated: 'As subcontractors for Google we have been a canary in the AI coalmine calling out the precarious labor conditions we face being the human workers standing between large language models and their end users'.260 The President of AWU observed that this agreement was a victory that underscored 'the importance of solidarity across the artificial barriers which corporations increasingly rely on to lower labor standards and reduce worker power'.261 255 Eidelson, J. & Alba, D. (Nov 6, 2023) Google Content Writers at Accenture Vote to Join Union. Bloomberg. https://www.bloomberg. com/news/articles/2023-11-06/google-tech-support-contract-workers-vote-to-join-union 256 Clayton, J. (2021) The woman who took on Google and won. BBC, April 7, 2021. https://www.bbc.com/news/technology-56659212 257 Davis, P.M. (April 30, 2024) Contract tech workers unionize in hopes of job security in a precarious industry. Texas Standard. https:// www.texasstandard.org/stories/contract-tech-workers-unionize-in-hopes-of-job-security-in-a-precarious-industry/ 258 Council, S. (March 5, 2024) Second group of contracted Google workers laid off after forming union. SFGate. https://www.sfgate.com/ tech/article/youtube-music-union-google-layoff-18703954.php 259 Alphabet Workers Union, CWA Local 9009 (Dec 18, 2024) Google Help Workers Ratify Contract With Accenture, First in Alphabet Workers Union-CWA History. https://www.alphabetworkersunion.org/press/google-help-workers-ratify-collective-contract-with-accenture-firstin-alphabet-workers-union-cwa-history 260 Taylor, J. (Jan 23, 2024) Precarious conditions of AI ‘ghost workers’ revealed by Google termination of Appen contract, union says. The Guardian. https://www.theguardian.com/australia-news/2024/jan/23/precarious-conditions-of-ai-ghost-workers-revealed-by-googletermination-of-appen-contract-union-says 261 Alphabet Workers Union, CWA Local 9009 (Dec 18, 2024) Google Help Workers Ratify Contract With Accenture, First in Alphabet Workers Union-CWA History. https://www.alphabetworkersunion.org/press/google-help-workers-ratify-collective-contract-with-accenture-firstin-alphabet-workers-union-cwa-history 61 ILO Working Paper 144 A more recent campaign has focused on a change in Google’s policy to mandate a minimum wage of at least $15 at its suppliers and contractors.262 AWU attributes this change to concerns with avoiding joint employer liability, as their interference in wage setting practices could be used to argue that they are a ‘joint employer’ in this case. AWU is organizing to demand improvements in pay and benefits for these workers - focusing on quality raters for Google Search, Gemini, and other Google products. Another campaign focuses on winning pay parity for all Temps, Vendors, and Contractors. The Alphabet Workers Union case is an example of solidarity among tech workers, with explicit support for improving conditions across fissured AI value chains. It shows the importance of both bottom-up organizing and solidarity from workers with greater institutional power - and also illustrates the way in which insecure contract employment is expanding alongside more permanent jobs within the same ‘Global North’ country. Alphabet, together with other leading tech firms, also subcontracts work in its AI value chain to lower wage countries in the Global South. In Case study 5 below, we examine a case of ‘bottom up’ labor organizing in Kenya, through the Kenyan Content Moderators Union. XCase study 5: Embedding AI labor in Africa: the Kenyan Content Moderators Union With the rise in AI models, demand has increased for 'data janitorial services' to structure and classify large amounts of data.263 The work is labor intensive and requires human intelligence. Workers identify and tag objects in images, or label audio and video messages known as data annotation. Other tasks include content moderation and matching information for better search engine optimization. Companies such as OpenAI and Meta are outsourcing their data annotation projects to firms such as Sama and IMerit that hire people in Global South countries like Kenya, South Africa and India. OpenAI had contracted with Sama to label tens of thousands of snippets of text that sometimes described disturbing situations in graphic detail. This work would contribute to a tool OpenAI was building to detect toxic content, which was eventually built into ChatGPT.264 Journalistic coverage of this kind of work has found that workers often experience long term mental health issues and report anger or anxiety responses due to constant exposure to disturbing images. Some workers quit due to suffering from post-traumatic stress disorder. Many are unable to receive an official diagnosis due to the inability to afford mental health care. While Sama offers on site counselling services, workers do not trust counselors or have been refused breaks. In response to these poor working conditions, many Sama workers quit in 2019. Others attempted to unionize. Early Strikes and Legal Action. In the summer of 2019, content moderators at Sama threatened to strike unless they were given better pay and conditions within seven days. The company responded by firing one of the leaders of the strike, Daniel Motaung, who was attempting to register the union. The company told other participating workers that 262 Allen, T. (2024) I’m paid $14 an hour to rate AI-generated Google search results. Subcontractors like me do key work but don’t get fair wages or benefits. Fortune, May 3, 2024. https://fortune.com/2024/05/03/google-search-raters-wages-benefits-contractors-tech-aiemployment/ 263 Bischoff, C., Kamoche, K., & Wood, G. (2024). The formal and informal regulation of labor in AI: The experience of Eastern and Southern Africa. ILR Review, 77(5), 825–835. 264 Perrigo, Billy. (2023, January 18). OpenAI Used Kenyan Workers on Less Than $2 Per Hour: Exclusive. Time. https://time.com/6247678/ openai-chatgpt-kenya-workers/ 62 ILO Working Paper 144 they could quit if they felt the conditions were unsatisfactory. The strike ended and there was no pay increase.265 However, Motaung sued Facebook and Samasource Kenya EPZ Ltd, for alleged workers’ rights violations, including exploitation, union busting and pay discrimination. Motaung claimed the work violated his basic human rights due to mental health and exposure to depictions of violence. Meta challenged the court’s jurisdiction in this case, saying it could not be a party since it was not registered in Kenya. This appeal was struck down by the Kenyan Employment and Labor Relations court in 2023, however, which found Meta could be sued in Kenya.266 In a related case, 183 content moderators sued Meta and its contractor Sama over unlawful and unfair termination of employment contracts. The Kenyan Human Rights Commission, the Ministry of Labor and the Central Organization of Trade Unions were named as interested parties. The workers alleged that they had been engaged by Meta and were recruited to work for the company by Sama; and that they were terminated because Motaung filed a constitutional petition challenging the gross violation of the moderator’s rights. The petition is unprecedented. It makes a case for holding Sama responsible for labor violations, but also Facebook and its parent company Meta, as they are in reality the employer of the moderators. The petition explains that Meta is 'in charge of all factors of production including Content Moderation which takes place at the workplace […]. The digital workplace is fitted with Surveillance features which keep track of all the work done, the time spent on each assignment and movements of the Facebook Content Moderators. The amount of time spent at the digital workplace together with other metrics is computed and billed for each content moderator. Their performance is reflected on their payslip which shows there is a direct connection between their work and their pay.'267 The petition asked for orders declaring the terminations to be unlawful, and that the moderators were engaged through inadequate pay and working conditions. It seeks compensatory damages worth $1.6 Billion. In addition, the moderators demanded that Meta cover the lifelong costs of mental health services for any mental health issues moderators may have developed from their work. Forming the African Content Moderators Union: The petition demanded that Meta and Sama recognize the workers’ right to form a union and talk about their work. Mophat Okinyi, the Chairperson of the African Content Moderators Union since 2023, observed, 'One of the biggest challenges has been the use of non-disclosure agreements (NDAs) and out-of-court settlements that silence workers from sharing their experiences. Additionally, the lack of recognition for unions by some tech companies and the dispersion of remote workforces make organizing and negotiating more difficult.'268 In May 2023, 150 content moderators voted at a landmark Nairobi meeting to register a Content Moderators Union.269 Initially, the union’s attempt to register was rejected by 265 Perrigo, Billy. (2022, May 11). Meta Accused Of Human Trafficking and Union-Busting in Kenya. Time, May 11, 2022. https://time. com/6175026/facebook-sama-kenya-lawsuit/ 266 Petition E071 of 2022. (2023, February 6). Kenya Law. https://kenyalaw.org/caselaw/cases/view/250879/ 267 Constitutional Petition E052 of 2023, 2023 268 Mophat Okinyi, Email Interview, December 22, 2024 269 Perrigo, Billy. (2023, May 1). 150 African Workers for AI Companies Vote to Unionize. Time. https://time.com/6275995/chatgpt-facebookafrican-workers-union/ 63 ILO Working Paper 144 the Labor Department since there were other collective bargaining bodies that represented the same kind of workers. In Kenya, two unions are not allowed to represent the interests of the same industry of workers. To solve this issue, the ACMU allied with the Communication Workers Union of Kenya (COWU-K) and is transforming into the African Content Moderators Association. It represents about 300 workers across contractors like Teleperformance (previously Majorel), Sama, Cloud Factory, NextState Foundation (called Stepwise in Kenya) and CCI. COWU-K is a traditional labor union, with established organizing strategies. Through this partnership, the Content Moderators Union is also educating established union leadership about new jobs like content moderation. However, the union has not negotiated collective agreements at the time of writing. Okinyi has also been part of developing other civil society platforms that help support the Content Moderators Union. He is the founder of Techworkers Community Africa, which educates young people on the benefits and threats of AI at the workplace. He is also involved in initiatives with the Partnership on AI, UNI Global Union, Solidarity Center, Siasa Place, and other advocacy groups working to create ethical standards for AI deployment. These initiatives aim to influence policies at national and international levels, including pushing for transparency, fair compensation, and protections against algorithmic exploitation. Okinyi observed that the success of these campaigns was built through both bottom up organizing and building coalitions with labor unions with organizational capacity at national and international level: 'Building trust and advancing solidarity were key strategies. I used worker-to-worker communication, storytelling, and highlighting shared challenges to create a sense of unity. Utilizing digital platforms to connect dispersed workers and partnering with organizations like the Communication Workers Union of Kenya and UNI Global Union helped to build collective power and provide resources for organizing.'270 He also outlined the union’s broader goals: 'Unions should advocate for worker representation in the design and governance of AI systems to ensure fair and ethical practices. They should also educate workers about their rights in AI-driven workplaces and push for legislative frameworks that protect labor rights in the digital age.' This case demonstrates the challenges faced by data workers and unions who are seeking to establish social dialogue with their employers, in fissured AI value chains that involve unequal power relationships between lead tech firms and their subcontractors. Their organizing and legal campaigns are helping to raise public awareness and pressure to improve conditions, which organizers hoped will lead to more formal collective bargaining in the longer term. These two cases illustrate both possibilities for and challenges to strengthening collective worker voice and building inclusive solidarity in the rapidly growing AI industry. The industry is organized by and through major tech firms. But employment conditions are determined by a decentralized network of contractors and platforms that face strong pressures to cut costs, necessary to compete for and keep contracts. The focus of unions’ organizing and social dialogue efforts 270 Mophat Okinyi, Email Interview, 22 December 2024 64 ILO Working Paper 144 has been on improving conditions across these fissured AI value chains through re-embedding jobs in labor and social protections - with some (if still limited) successes. 2.4. Summary Labor unions and other worker representatives have sought to influence employer decisions concerning how they deploy AI and algorithms in the workplace through a range of social dialogue and organizing strategies. Our analysis above has divided these activities into three categories, based on ‘action fields’ with distinct social dialogue topics and goals: skills and employment (from labor replacing to complementing); algorithmic management and monitoring (from labor controlling to empowering); and working conditions and rights in AI-enabled fissuring (from labor displacing to embedding). Case study findings show different patterns of social dialogue practices and outcomes across world regions and countries. We find the largest number of examples in Europe of more formal, organized social dialogue through collective bargaining across these three action fields. Unions and works councils have negotiated comprehensive agreements establishing job security or dismissal protection rights, strengthening investments in skills and training, and articulating clear guidelines concerning how employers use AI tools. They have used stronger data protection and co-determination or bargaining rights to win broad provisions limiting the use of algorithmic management software for worker surveillance and establishing a ‘humanin-command’ approach in a growing number of industries and firms. AI ethics committees and other joint labor-management committees are also being established to oversee the fairness of algorithmic management tools ─ most notably in Germany. Both union organizing and social dialogue institutions have been mobilized to counter the use of untransparent algorithms to direct and evaluate workers. At the same time, the focus of agreements differs significantly across the European countries we have examined. Nordic unions have the most organized social dialogue over ‘labor complementing’ uses of AI, and there is more widespread emphasis on limiting ‘labor control’ via algorithmic management in France, Italy, and Spain. The German case shows the most comprehensive agreements in both areas, with unions and works councils seeking to pursue a combination of partnership-based investments in productivity improvement complemented by skill investments and limits on algorithmic control - mobilizing strong co-determination rights in this area. Despite their weaker labor laws, bargaining rights, and data protection rules, US and Canadian unions have also engaged in creative social dialogue on AI with employers, negotiating innovative agreements that encourage employers to invest in worker skills and discretion, while restricting invasive monitoring and control through algorithms. Creative workers - artists, writers, and actors - have won original protections that address worker control over decisions concerning how their work is complemented or replaced by generative AI, including ownership of their own images and art. While the European cases we have discussed in this report often involved a combination of conflict and cooperation, in many of the US cases, unions organized more protracted strikes or threatened industrial action to win AI protections. The case of tech worker organizing in the US shows a distinctive example of pursuing social dialogue outside of formal collective bargaining - including within Alphabet, one of the lead firms developing and deploying AI. Between our two main East Asian cases, we observe an interesting contrast between predominantly cooperative social dialogue in Japan, focusing on productivity improvements and upskilling in a context of structural labor shortages, and labor militancy in South Korea, focused on reinstating workers laid off due to automation. However, in the area of algorithmic management, 65 ILO Working Paper 144 case studies in both countries show some degree of labor conflict, resolved through worker mobilization or legal challenges. In India, Brazil, the Dominican Republic, and Kenya, social dialogue on AI topics was more nascent. In Brazil, established unions were organizing broader projects focused on regulating AI and encouraging skill upgrading in traditional industries such as banking - where gender equity was also a focus. One of the most visible of these have focused on addressing precarious and unsafe conditions associated with AI-based fissuring in data labeling and content moderation work in Kenya. Overall, our findings from this comparison of social dialogue over AI at industry, company, and workplace-levels demonstrate a range of creative strategies and agreements across the three action fields. These seek to encourage more socially beneficial and economically just applications of these new technologies - but also to ensure that they are produced and deployed under fair conditions. Social dialogue on AI has been most successful in achieving these goals where it institutionalizes protections in new laws, policies, and collective agreements or is grounded in an existing framework of institutionalized protections. However, newer organizing and mobilization efforts are an important first step to establishing these more encompassing institutions. 66 ILO Working Paper 144 XConclusion In this report, we have presented case studies of social dialogue over AI and algorithmic management in different countries and world regions. We attempted to identify cases that represent a cross-section of global developments; but our analysis is by no means comprehensive or complete. It does, however, illustrate the important role that social dialogue is playing internationally in supporting the transition to a more just and ethical ‘digital economy’. In this conclusion, we attempt to draw some broader lessons concerning the conditions that support effective social dialogue across the three ‘action fields’ we have examined in our report: from labor replacing to complementing, labor controlling to empowering, and labor displacing to embedding. We return here to our framework, developed in Section 1.1 above, which outlined three factors or conditions that we argued play an important role in supporting more inclusive and effective social dialogue on AI: constraints on employer exit, support for collective worker voice, and strategies of inclusive solidarity.271 Figure 1 - reproduced below - illustrates the overlap of this framework with the three ‘action fields’ we have organized this report around. XFigure 1: Supporting social dialogue on AI through constraints on exit, support for voice, and strategies of solidarity The social dialogue examples we have discussed in this report can be seen as attempts to establish or strengthen constraints on employer exit and support for collective worker voice, while deploying more inclusive strategies of solidarity. While all three play an important role across 271 Doellgast, V. (2022). Exit, voice, and solidarity: Contesting precarity in the US and European telecommunications industries. Oxford University Press 67 ILO Working Paper 144 social dialogue and union campaigns, our case studies suggest that each action field relies most centrally on two of the three factors. First, social dialogue encouraging a shift from labor replacing to labor complementing uses of AI requires a combination of constraints on employer exit and support for collective worker voice. In the cases we reviewed, strong employment protections and skill investments were central goals - and these made it more difficult or less desirable for firms to exit their internal workforce. Laws establishing clear rules and copyright protections on the use of generative AI to reproduce art, voice, or images are one example. Collective agreements across countries provided job security, commitments to decrease subcontracting, or support for retraining and redeploying workers. The most widely publicized cases of negotiations in the US over AI use in film, television, and game development all involved clear negotiated constraints on companies’ ability to use AI to ‘exit’ from artists’ and writers’ past rights and contractual protections. In addition, successful social dialogue across these examples would not have been possible without strong support for collective worker voice. In Europe, this support came from strong participation rights and laws, as well as traditions of strong unions and tripartite social dialogue. Across countries, worker mobilization, in some cases via strikes, were crucial for winning strong agreements on job security or that limited how employers could use AI at work. Strengthening worker voice - or establishing its central role in AI decision-making - has also been a key demand of unions in social dialogue at national and European levels. And joint projects mapping needed investments in skills and training, such as in the Brazilian banking industry, as well as on increasing productivity while returning these gains to workers, as in Japan’s AEON agreements, certainly relied on worker voice via worker representatives to establish needs and opportunities. Second, social dialogue encouraging a shift from labor controlling to labor empowering uses of AI requires a combination of support for collective worker voice and strategies of inclusive labor solidarity. Institutions and laws supporting collective voice have been a central tool but also key labor demand. In Germany, for example, laws providing co-determination rights in cases where technology was used for performance and behavior control, or in Sweden, requirements that firms negotiate over technology’s impacts on health and safety, were crucial for successful social dialogue limiting the invasive use of algorithmic management and protecting worker discretion in the workplace. Across countries, and particularly in Europe and South Korea, data protection laws were important tools for contesting worker surveillance and for providing worker representatives with information on what data was being collected on workers or how it was used. And bright line prohibitions of certain uses of AI to automate HR decisions or use sentiment analysis tools in the workplace could support establishing more fair and transparent workplace rules. In addition, unions worldwide have sought to encourage provisions in laws and collective agreements granting worker representatives stronger rights to consult or negotiate over these tools. Strategies of inclusive labor solidarity are particularly important for regulating algorithmic management because these tools are more intensively and invasively deployed across a contracted or fissured workforce. We presented cases that showed how traditional unions and newer worker organizations have focused on limiting the use of algorithmic management and performance monitoring systems in outsourced BPO call centers and AI data coders and content moderators. The case of Hyundai Heavy Industries in South Korea shows the importance of solidarity between unions, in that case for challenging the discriminatory use of facial recognition for in-house subcontractors. Establishing better safeguards against bias in large language models, and their application in the workplace, also requires a solidaristic movement, centering on injustice for workers who also tend to hold more precarious contracts. 68 ILO Working Paper 144 Third, social dialogue encouraging a shift from labor displacing to labor embedding in AIenabled location and organizational changes was most successful where it combined constraints on employer exit with strategies of inclusive labor solidarity. In our discussion of the Alphabet Workers Union and the Kenyan Content Moderators Union, we saw the central importance of inclusive solidarity in these campaigns. Labor solidarity was necessary to build a strong movement of tech workers in precarious jobs at the bottom of the AI value chain. It also was important in extending bargaining power from more protected tech professionals, to build new agreements and institutions that constrain employer exit from these protections through strengthening employment standards and protections for temps, vendors, and contractors. In sum, social dialogue can and is playing a crucial role in encouraging an alternative, high road approach to AI investments and uses in workplaces around the world. In the end, we find more similarities than differences across the case studies we have presented in this report: together they illustrate the shared goals of workers and their unions to move their employers and governments toward AI strategies that complement, empower, and embed their labor. Efforts to promote more socially and economically sustainable approaches to new technology adoption should focus on establishing complementary institutions and practices that are designed by the workers most directly affected by these changes. 69 ILO Working Paper 144 Annex 1. List of interviews and email communication Interviews Name and/or position Organization Country Date Works councilor 1 Deutsche Telekom Germany 14 August 2024 Works councilor 2 Deutsche Telekom Germany 14 August 2024 Works councilor IBM Germany 26 July 2024 Works councilor Pronvinzial Germany 24 October 2024 Franca Salis-Manidier, National secretary of the CFDT, responsible for European and international affairs, digital and artificial intelligence French Democratic Confederation of Labor (CFDT) France 13 January 2025 Principal Officer Teamsters Local 2 USA 5 November 2024 Researcher Culinary Union, UNITE HERE Local 226 USA 18 November 2024 Executive Vice President, General Counsel SAG-AFTRA USA 6 November 2024 Unit Chair Ziff-Davis Creators GuildWGA USA 7 November 2024 Jayson Little, Staff Representative United Steel Workers Canada 6 November 2024 Mr. Koki Ueyama, Chief Secretary (Interview conducted by Eriko Teramura) Federation of AEON Group Workers’ Unions Japan 26 December 2024 Representatives (Interview conducted by Eriko Teramura) Federation of AEON Group Workers’ Unions Japan 2 February 2025 Hyunju Kim, President of the DeunDeunHan Call Center Union DeunDeunHan Call Center Union South Korea 26 December 2024 Hanoi Sosa, organizer and secretary of Workers’ Education to Dominican federation of free-trade zone workers, and secretary-general to FEDOTRAZONAS. Dominican Republic 15 November 2024 76 ILO Working Paper 144 Feng, Kung. 2023. "Overview of New Rights for Workers Under the California Consumer Privacy Act." UC Berkeley Labor Center. https://laborcenter.berkeley.edu/overview-of-new-rights-forworkers-under-the-california-consumer-privacy-act/. Freeman, Richard B., and Kathryn L. Shaw. International Differences in the Business Practices and Productivity of Firms. Chicago: University of Chicago Press, 2009. Frey, Carl Benedikt, and Michael A. Osborne. "The Future of Employment: How Susceptible Are Jobs to Computerisation?" Technological Forecasting and Social Change 114 (2017): 254-280. Gerst, Detlef. "Geschäftsmodelle mitentwickeln-- ein neues Handlungsfeld der Betriebsräte." WSIMitteilungen 73 (2020): 295-299. doi: 10.5771/0342-300X-2020-4-295. Global Deal. "Social Dialogue and the Use of Artificial Intelligence in the Workplace." 2024. Goldman, Debbie J. Disconnected: Call Center Workers Fight for Good Jobs in the Digital Age. University of Illinois Press, 2024. Graham, Mark, ed. Digital Economies at Global Margins. Cambridge: MIT Press, 2019. Graham, Mark, Isis Hjorth, and Vili Lehdonvirta. "Digital Labor and Development: Impacts of Global Digital Labor Platforms and the Gig Economy on Worker Livelihoods." Transfer: European Review of Labor and Research 23, no. 2 (2017): 135-162. Gray, Mary L., and Siddharth Suri. Ghost Work: How to Stop Silicon Valley from Building a New Global Underclass. Boston: Eamon Dolan Books, 2019. Guaglianone, Luciana. "Collective Bargaining and AI in Italy." In Artificial Intelligence, Labour and Society, edited by Aída Ponce Del Castillo, 207-215. ETUI, 2024. Guillot, Alison. "The Culinary Academy of Las Vegas and Amadeus Join Forces to Expand Hospitality Employment and Job Training." Amadeus, 2024. https://www.amadeus-hospitality.com/insight/ the-culinary-academy-of-las-vegas-and-amadeus-join-forces-to-expand-hospitality-employmentand-job-training/. Guridi, José Antonio. "Dominican Republic: Artificial Intelligence Readiness Assessment Report." UNESCO. 2024. Habuka, Hiroki. "Japan's Approach to AI Regulation and Its Impact on the 2023 G7 Presidency." CSIS, February 14, 2023. https://www.csis.org/analysis/japans-approach-ai-regulation-and-itsimpact-2023-g7-presidency. Haipeter, Thomas, Manfred Wannöffel, Jan-Torge Daus, and Sandra Schaffarczik "Human-Centered AI through Employee Participation." Frontiers in Artificial Intelligence 7 (2024). Hall, Peter A., and David Soskice, eds. Varieties of Capitalism: The Institutional Foundations of Comparative Advantage. Oxford: Oxford University Press, 2001. Harbecke, Tim, and Gernot Mühge. "Digitalisierungsstrategien im Portrait." 34, Mitbestimmungspraxis. Düsseldorf: Hans Böckler Foundation, 2020. 77 ILO Working Paper 144 Hays, Seth. 2024. "AI Training and Copyright Infringement: Solutions from Asia." Tech Policy Press, October 30, 2024. https://www.techpolicy.press/ai-training-and-copyright-infringementsolutions-from-asia/. IATSE. "Summary of Basic Agreement Negotiations." 2024. https://iatse.net/wp-content/ uploads/2024/06/2024-SUMMARY-OF-BASIC-AGREEMENT-NEGOTIATIONS_6.28.24-FINAL.pdf. ILO. "Social Dialogue." 2019. https://www.ilo.org/resource/social-dialogue-0. Ilsøe, Anna, Trine P. Larsen, Christopher Mathieu, and Bertil Rolandsson. "Negotiating about Algorithms: Social Partner Responses to AI in Denmark and Sweden." ILR Review 77, no. 5 (2024): 856-868. Inditex. "Joint Statement by the Inditex Group and the European Works Council about the Digital Transformation Strategic Plan." 2020. https://www.uni-europa.org/old-uploads/2020/12/ EN\_-DECLARACION-CONJUNTA-INDITEX-Y-CEE-ANTE-EL-PLAN-ESTRAT%C3%89GICO-DETRANSFORMACI%C3%93N-DIGITAL.pdf. IndustriALL. "IndustriALL Adopts a Trade Union Strategy to Tackle AI at Work." December 4, 2024. https://news.industriall-europe.eu/Article/1176. Innovation, Science and Economic Development Canada. "The Artificial Intelligence and Data Act (AIDA) -- Companion Document." March 13, 2023a. https://ised-isde.canada.ca/site/innovationbetter-canada/en/artificial-intelligence-and-data-act-aida-companion-document. Innovation, Science and Economic Development Canada. "Voluntary Code of Conduct on the Responsible Development and Management of Advanced Generative AI Systems." September 2023b. https://ised-isde.canada.ca/site/ised/en/voluntary-code-conduct-responsible-developmentand-management-advanced-generative-ai-systems. Digital Policy Alert. Japan: Closed Consultation on AI Business Guidelines. 2024. https:// digitalpolicyalert.org/event/16812-closed-consultation-on-ai-operator-guidelines. JMIU日本アイビーエム支部. "──不当労働行為事件勝利和解." JMITU日本アイビーエム支部, 2024. https://www.jmitu-ibm.org/2024/08/9302.html. Johnson, Simon, and Daron Acemoglu. Power and Progress: Our Thousand-Year Struggle over Technology and Prosperity. London: Hachette UK, 2023. Joint Declaration on Employment Aspects of Artificial Intelligence by the European Social Partners in the Banking Sector. May 14, 2024. https://www.uni-europa.org/wp-content/uploads/ sites/3/2024/05/Joint-Declaration-on-Employment-Aspects-of-Artificial-Intelligence_BankSD.pdf. Kageni, Mary, and Yvonne Odhiambo. "Strengthening Data Protection in Kenya: Opportunities and the Way Forward -- KIPPRA." n.d. Accessed January 14, 2025. https://kippra.or.ke/strengtheningdata-protection-in-kenya-opportunities-and-the-way-forward/. Kämpf, Tobias, Barbara Langes, Lars Christian Schatilow, and Hans-Joachim Gergs, eds. Human Friendly Automation: Arbeit und Künstliche Intelligenz neu denken. Frankfurter Allgemeine Buch, 2023. 78 ILO Working Paper 144 Kang, Cecilia. "States Take Up AI Regulation Amid Federal Standstill." New York Times, June 10, 2024. https://www.nytimes.com/2024/06/10/technology/california-ai-regulation.html. Kassem, Sarrah. Work and Alienation in the Platform Economy: Amazon and the Power of Organization. Bristol: Policy Press, 2023. Khan, Mishal, Annette Bernhardt, and Laura Pathak. "The Current Landscape of Tech and Work Policy: A Guide to Key Concepts." 2024. https://laborcenter.berkeley.edu/tech-and-work-policyguide/. Kim, Geun-Ju. "Court Finds Union's Argument That Facial Recognition Devices May Violate Personal Information Valid." Yonhap News, 2024. https://www.yna.co.kr/view/AKR20240704114000057. Kim, Sunghoon, and Seri No. “Workplace AI Regulation and Worker Resistance in Korea.” ILR Review 77, no. 5 (2024): 835–846. Kim, Soojong, Poong Oh, and Joomi Lee. "Algorithmic Gender Bias: Investigating Perceptions of Discrimination in Automated Decision-making." Behaviour & Information Technology 43, no. 16 (2024): 4208-4221. Kitagawa, Keiichi. "IBM Japan to Disclose How AI is Used to Evaluate Its Work Force." The Asahi Shimbun, 2024. https://www.asahi.com/ajw/articles/15375025. Knotter, Ad. Transformations of Trade Unionism: Comparative and Transnational Perspectives on Workers Organizing in Europe and the United States. Amsterdam University Press, 2018. Kochan, Thomas A., and Paul Osterman. The Mutual Gains Enterprise. Cambridge: Harvard Business School Press, 1994. Korean Confederation of Trade Unions. "AI Framework Legislation: Industrial Development Is Not Enough—The Need for Thorough Impact Assessments and Remedies for Labor and Human Rights Risks." Press release, October 31, 2024. https://nodong.org/statement/7869923. Kordzadeh, Nima, and Maryam Ghasemaghaei. "Algorithmic Bias: Review, Synthesis, and Future Research Directions." European Journal of Information Systems 31, no. 3 (2022): 388-409. Kresge, Lisa. "Negotiating Workers' Rights at the Frontier of Digital Workplace Technologies in 2023." UC Berkeley Labor Center, 2023. Kresge, Lisa. "Union Collective Bargaining Agreement Strategies in Response to Technology." Working Paper, Technology and Work Program. Center for Labor Research and Education, University of California, Berkeley, November 2020. Krzywdzinski, Martin. "Zwei Welten der KI in der Arbeitswelt Wie Management und Betriebsräte die Einführung und Nutzung von KI-Anwendungen gestalten." Weizenbaum Discussion Paper #39, June 2024. https://www.weizenbaum-library.de/server/api/core/bitstreams/dd69f1ed-98ce-4b64a996-d44521919ba3/content. Krzywdzinski, Martin, Daniel Schneiß, and Andrea Sperling. "Between Control and Participation: The Politics of Algorithmic Management." New Technology, Work and Employment 40, no.1 (2025): 60-80. 79 ILO Working Paper 144 Krzywdzinski, Martin, Detlef Gerst, and Florian Butollo. "Promoting Human-Centred AI in the Workplace. Trade Unions and Their Strategies for Regulating the Use of AI in Germany." Transfer: European Review of Labour and Research 29, no. 1 (2023): 53-70. Kulkarni, Pranav. "South African Labor Court Interdicts Country-Wide Strike by Bank Officials." People's Dispatch, 2019. https://peoplesdispatch.org/2019/09/27/south-african-laborcourt-interdicts-country-wide-strikeby-bank-officials/. Le Bonniec, Tom. 2024. "Another Path for AI Regulation: Worker Unions and Data Protection Rights." Italian Labour Law e-Journal 17, no. 2: 115-131. https://doi.org/10.6092/issn.15618048/20870. Lee, Jongho. "Introducing CCTV and Facial Recognition Devices is an Illegal Human Rights Violation -- Union Files Complaint Against HD Hyundai Heavy Industries." Ulsan Journal, 2024. https://m. usjournal.kr/news/newsview.php?ncode=1065589840144988. Litwin, Adam Seth, and Gali Racabi. "Varieties of AI Regulations: The United States Perspective." ILR Review 77, no. 5 (2024): 799-812. Liu, Mingwei, Hao Zhang, and Yi Sui. "Workplace Artificial Intelligence Regulation in China: Between Efficiency and Social Stability." ILR Review 77, no. 5 (2024): 813-824. Local Progress and in the Public Interest. "Harnessing the Power of Procurement: Issues, Considerations, and Best Practices to Advance Equity in the Contracting of Public Goods and Services." 2023. https://localprogress.org/2023/07/12/harnessing-the-power-of-procurement/. Maake, Sydney Rapelo. Exploring SASBO'S Response to the Challenges of Artificial Intelligence in the Banking Industry in Gauteng. Thesis, University of Johannesburg, 2022. https://hdl.handle. net/10210/499894. Maddaus, Gene. "SAG-AFTRA Wins Passage of California Bill to Limit AI Replicas." Variety, 2024. https:// variety.com/2024/politics/news/sag-aftra-limit-ai-digital-replica-california-bill-1236120379/. Master Agreement between General Motors of Canada Company and UNIFOR, Memorandum of Understanding Health and Safety. November 5, 2020. https://negotech.service.canada.ca/eng/ agreements/01/0194612a.pdf. McElheran, Kristina, J. Frank Li, Erik Brynjolfsson, Zachary Kroff, Emin Dinlersoz, Lucia Foster, and Nikolas Zolas. "AI Adoption in America: Who, What, and Where." Journal of Economics & Management Strategy 33, no. 2 (2024): 375-415. Memorandum of Agreement between the News Media Guild and the Associated Press. 2024. https://newsmediaguild.org/wp-content/uploads/2024/05/Contract-2024-Editorial-Unit.pdf. METI. "Governance Guidelines for Implementation of AI Principles." January 28, 2022. https:// www.meti.go.jp/shingikai/mono_info_service/ai_shakai_jisso/pdf/20220128_2.pdf. Metz, Cade. "A.I. Is Learning From Humans. Many Humans." The New York Times, August 16, 2019. https://www.nytimes.com/2019/08/16/technology/ai-humans.htm. 80 ILO Working Paper 144 Mgala, Mvurya. "The Extent and Use of Artificial Intelligence to Achieve the Big Four Agenda in Kenya." Multidisciplinary Journal of Technical University of Mombasa 1, no. 1 (2020): 1-7. Ministry of Labor and Social Economy. "BOE-A-2023-12870 Resolución de 19 de mayo de 2023, de la Dirección General de Trabajo, por la que se registra y publica el V Acuerdo para el Empleo y la Negociación Colectiva." Boletín Oficial del Estado, no. 129, May 31, 2023: 75426–75447. https:// www.boe.es/diario_boe/txt.php?id=BOE-A-2023-12870 Mohanty, Amlan, and Shatakratu Sahu. "India's Advance on AI Regulation." Carnegie Endowment for International Peace, November 21, 2024. https://carnegieendowment.org/research/2024/11/ indias-advance-on-ai-regulation?lang=en¢er=india. Muldoon, James, Mark Graham, and Callum Cant. Feeding the Machine: The Hidden Human Labour Powering AI. Edinburgh: Canongate Books, 2024. National Nurses United. "A.I.'s Impact on Nursing and Health Care." 2024. https://www. nationalnursesunited.org/artificial-intelligence. National Nurses United. "National Nurses United Survey Finds A.I. Technology Degrades and Undermines Patient Safety." 2024. https://www.nationalnursesunited.org/press/national-nursesunited-survey-finds-ai-technology-undermines-patient-safety. National Nurses United. "Nurses and Patients' Bill of Rights: Guiding Principles for AI Justice in Nursing and Health Care." 2024. https://www.nationalnursesunited.org/sites/default/files/nnu/ documents/0424_NursesPatients-BillOfRights_Principles-AI-Justice_flyer.pdf. National Nurses United. "Nurses March Nationwide in Support of Safe Staffing, Patient Protections Against A.I." January 14, 2025. https://www.nationalnursesunited.org/press/nurses-marchnationwide-in-support-of-safe-staffing-and-patient-protections-against-ai. NBA-NBPA Collective Bargaining Agreement. 2017. https://cosmic-s3.imgix.net/3c7a0a50-8e1111e9-875d-3d44e94ae33f-2017-NBA-NBPA-Collective-Bargaining-Agreement.pdf. New York News Guild. "Prime Day Walkout Averted at Ziff Davis." July 15, 2024. https://newsguild. org/prime-day-walkout-averted-at-ziff-davis/. New York State Senate. Senate Bill S7623: An Act to Amend the Labor Law, in Relation to Restricting the Use of Electronic Monitoring and Automated Employment Decision Tools. 2023–2024 Regular Sessions. Introduced August 4, 2023. https://legislation.nysenate.gov/pdf/bills/2023/S7623. NITI. "Responsible AI Approach Document for India: Part 2 -- Operationalizing Principles for Responsible AI." August 2021. https://www.niti.gov.in/sites/default/files/2021-08/Part2Responsible-AI-12082021.pdf. O'Brady, Sean, and Virginia Doellgast. "Collective Voice and Worker Well‐being: Union Influence on Performance Monitoring and Emotional Exhaustion in Call Centers." Industrial Relations: A Journal of Economy and Society 60, no. 3 (2021): 307-337. OECD. OECD Employment Outlook 2023: Artificial Intelligence and the Labour Market. Paris: OECD Publishing, 2023. https://doi.org/10.1787/08785bba-en. 81 ILO Working Paper 144 Office of Management and Budget. "Memorandum of Advancing Governance, Innovation, and Risk Management for Agency Use of Artificial Intelligence." 2024. https://www.whitehouse.gov/ wp-content/uploads/2024/03/M-24-10-Advancing-Governance-Innovation-and-Risk-Managementfor-Agency-Use-of-Artificial-Intelligence.pdf. Okolo, Chinasa T. "Reforming Data Regulation to Advance AI Governance in Africa." Brookings, March 15, 2024. https://www.brookings.edu/articles/reforming-data-regulation-to-advance-aigovernance-in-africa/. Orange. "Orange Adopts a Data and Artificial Intelligence Ethical Charter." November 15, 2022. https://newsroom.orange.com/orange-adopts-a-data-and-artificial-intelligence-code-of-ethics/. Özkiziltan, Didem. "Governing Engels' Pause: AI and the World of Work in Germany." ILR Review 77, no. 5 (2024): 846-856. Özkiziltan, Didem and Fabio Landini, “Trustworthy and Human-Centric? The New Governance of Workplace AI Technologies under the EU's Artificial Intelligence Act,” Transfer, no. 4 (2025). Parliament of Canada. Bill C-27: Digital Charter Implementation Act, 2022. 2022. https://www. parl.ca/legisinfo/en/bill/44-1/c-27. Partnership on AI. "Responsible Sourcing of Data Enrichment Services." 2021. http://partnershiponai. org/responsible-sourcing. Perold, Colette, and Eric Dirnbach. "What Is Pre-Majority Unionism?" 2025. https://workerorganizing. org/premajority-unionism/defined/. Perrigo, Billy. "150 African Workers for AI Companies Vote to Unionize." Time, May 1, 2023. https:// time.com/6275995/chatgpt-facebook-african-workers-union/. Perrigo, Billy. "Meta Accused of Human Trafficking and Union-Busting in Kenya." Time, May 11, 2022. https://time.com/6175026/facebook-sama-kenya-lawsuit/ Perrigo, Billy. "OpenAI Used Kenyan Workers on Less Than $2 Per Hour: Exclusive." Time, January 18, 2023. https://time.com/6247678/openai-chatgpt-kenya-workers/ Personal Information Protection Commission. "Amendments to the Personal Information Protection Act in 2023: What's Changed?" Korea Policy Briefing, December 29, 2023. https:// www.korea.kr/briefing/pressReleaseView.do?newsId=156608465 Petition E071 of 2022. Kenya Law, February 6, 2023. https://kenyalaw.org/caselaw/cases/ view/250879/ Pfeffer, Jeffrey. Dying for a Paycheck: How Modern Management Harms Employee Health and Company Performance---and What We Can Do About It. New York: HarperBusiness, 2018. Ponce Del Castillo, Aída. "AI and Trade Unions: From Rapid Responses to Proactive Strategies." HesaMag 29 (Winter 2024): 31-34. Professional Institute of the Public Service of Canada. "Brief to the House of Commons Standing Committee on Industry and Technology (INDU) On Bill C-27." August 2023. https://www. 82 ILO Working Paper 144 ourcommons.ca/Content/Committee/441/INDU/Brief/BR12565509/br-external/ProfessionalIn stituteOfThePublicServiceOfCanada-e.pdf Progressive Network Center & People's Solidarity for Participatory Democracy. "Joint Commentary: Regret Over the Passage of the AI Framework Act Neglecting Human Rights." December 27, 2024. https://www.peoplepower21.org/publiclaw/1983459. Rani, Uma, Morgan Williams, and Nora Gobel. The Human Cogs in the AI Machine: Experiences of Data Annotation and Content Moderation Workers in the BPO Sector in India and Kenya. ILO Working Paper. Forthcoming. Regeringskansliet. "AI-kommissionens Färdplan för Sverige." 2024. https://www.regeringen.se/ rapporter/2024/11/ai-kommissionens-fardplan-for-sverige/. Rho, Hye Jin, and Christine Riordan. "Beyond Algorithmic Control: Exploring Self-Sequencing of Tasks in Hotel Housekeeping Work." Draft paper presented at the 'AI and the Future of Work Conference', Cornell University, October 11-18, 2024. Rolf, Steven. "AI and Algorithmic Management in European Services Sectors: Prevalence, Functions, and a Guide for Negotiators." UNI Europa and Friedrich Ebert Stiftung, 2024. Romm, Tony. "Inside the Lobbying War Over California's Landmark Privacy Legislation." The Washington Post, February 8, 2019. https://www.washingtonpost.com/technology/2019/02/08/ theres-going-be-fight-here-weaken-it-inside-lobbying-war-over-californias-landmark-privacy-law/. Ross, Philip. "Waterfront Labor Response to Technological Change: A Tale of Two Unions." Labor Law Journal 21, no. 7 (1970): 397. SAG-AFTRA. "Comparison of Union and Employer Proposals on Artificial Intelligence." 2024. https:// www.sagaftra.org/sites/default/files/sa_documents/ima_comparison_chart.pdf. SAG-AFTRA. "New SAG-AFTRA and Ethovox Agreement Empowers Actors and Secures Essential A.I. Guardrails." October 28, 2024. https://www.sagaftra.org/new-sag-aftra-and-ethovox-agreementempowers-actors-and-secures-essential-ai-guardrails. Sasbo. “Why Sasbo is Challenging the Banks on 4IR Issues.” Sasbo News, vol. 41, no. 4, Oct./ Nov. 2019. https://www.sasbo.org.za/wp-content/uploads/2020/04/SASBO-News-v41n4-Proof27-Nov.pdf. Schulte, Paul A., Steven L. Sauter, Sudha P. Pandalai, Hope M. Tiesman, Lewis C. Chosewood, Thomas R. Cunningham, Steven J. Wurzelbacher et al. "An Urgent Call to Address Work‐Related Psychosocial Hazards and Improve Worker Well‐Being." American Journal of Industrial Medicine 67, no. 6 (2024): 499-514. SindBancarios. "Mais mulheres na TI: abertas inscrições para cursos reivindicados pelo movimento sindical." March 12, 2025. https://www.sindbancariospetropolis.com.br/noticia/mais-mulheresna-ti-abertas-inscricoes-para-cursos-reivindicados-pelo-movimento-sindical/12103. Spilda, Funda Ustek, Lola Brittain, Callum Cant, and Mark Graham. "The Unmagical World of AI: Workers at the Bottom of the AI Supply Chain." Friedrich Ebert Stiftung, February 26, 2024. https://futureofwork.fes.de/news-list/e/ai-value-chain.html 83 ILO Working Paper 144 Stanek, Thomas M, and Zachary V. Zagger. "NLRB Acting General Council Rescinds Many of Predecessor's Memos, Sets Stage for New Labor Policy." The National Law Review, February 15, 2025. https://natlawreview.com/article/nlrb-acting-general-counsel-rescinds-many-predecessorsmemos-sets-stage-new-labor. State of California. "GenAI Guidelines for Public Sector Procurement, Uses and Training." 2024. https://cdt.ca.gov/wp-content/uploads/2024/07/3a-GenAI-Guidelines.pdf. Stephan, Katcy. "SAG-AFTRA Video Game Performers Hold Picket Outside WB Games as Strike Continues: 'We Are Still Fighting'." Variety, March 5, 2025. https://variety.com/2025/biz/news/sagaftra-video-game-strike-picket-1236328677/. Streeck, Wolfgang. 1992. "Productive Constraints: On the Institutional Conditions of Diversified Quality Production." Social Institutions and Economic Performance: Studies of Industrial Relations in Advanced Capitalist Economies, 1-40. Susskind, Daniel. A World Without Work: Technology, Automation and How We Should Respond. London: Penguin UK, 2020. Tan, Jessica S., Nataliya Nedzhvetskaya, and Elizabeth Mazo. "Unlikely Organizers: The Rise of Labor Activism Among Professionals in the US Technology Industry." Working Paper, 2023. Tarifvertrag zur Zukunft der Arbeit im Rahmen der Digitalisierung im DB-Konzern (TV Arbeit 4.0 EVG 2018). Berlin/Frankfurt am Main, December 14, 2018.https://www.evg-online.org/fileadmin/ Tarif/Tarifvertraege/Tarifvertraege_DB_Konzern/TV_Arbeit_4.0_EVG_2018_internet.pdf. Taylor, Josh. "Precarious Conditions of AI 'Ghost Workers' Revealed by Google Termination of Appen Contract, Union Says." The Guardian, January 23, 2024. https://www.theguardian.com/australianews/2024/jan/23/precarious-conditions-of-ai-ghost-workers-revealed-by-google-termination-ofappen-contract-union-says. Telecom Social Dialogue Committee. "Joint Declaration on Artificial Intelligence." 2020. https:// www.uni-europa.org/old-uploads/2020/12/20201130_UE-ETNO-declaration-AI.pdf. The White House. "Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence." October 30, 2023. https://www.whitehouse.gov/briefing-room/presidentialactions/2023/10/30/executive-order-on-the-safe-secure-and-trustworthy-development-and-useof-artificial-intelligence/. The White House. "Readout of Vice President Harris's Meeting with Consumer Protection, Labor, and Civil Rights Leaders on AI." July 13, 2023. https://www.whitehouse.gov/briefing-room/statementsreleases/2023/07/13/readout-of-vice-president-harriss-meeting-with-consumer-protection-laborand-civil-rights-leaders-on-ai/. The White House. "Readout of White House Listening Session with Union Leaders on Advancing Responsible Artificial Intelligence Innovation." July 3, 2023. https://www.whitehouse.gov/briefingroom/statements-releases/2023/07/03/readout-of-white-house-listening-session-with-unionleaders-on-advancing-responsible-artificial-intelligence-innovation/. 84 ILO Working Paper 144 The White House. "Removing Barriers to American Leadership in Artificial Intelligence." January 23, 2025. https://www.whitehouse.gov/presidential-actions/2025/01/removing-barriers-to-americanleadership-in-artificial-intelligence/. Tubaro, Paola, Marion Coville, Clément Le Ludec, and Antonio A. Casilli. "Hidden Inequalities: The Gendered Labour of Women on Microtasking Platforms." Internet Policy Review 11, no. 1 (2022): 1-26. TUC. "Artificial Intelligence (Regulation and Employment Rights) Bill." 2024. https://www.tuc.org. uk/research-analysis/reports/artificial-intelligence-regulation-and-employment-rights-bill. UAW Ford Contract. 2023. https://uaw.org/2023fordcontract/vol-4a/#p=1. U.S. Government Accountability Office. “Digital Surveillance of Workers: Tools, Uses, and Stakeholder Perspectives.” GAO-24-107639. Washington, DC: U.S. Government Accountability Office, 2024. https://www.gao.gov/assets/gao-24-107639.pdf. U.S. Senate. “Statements from the Third Bipartisan Senate Forum on Artificial Intelligence.” November 6, 2023. https://www.schumer.senate.gov/newsroom/press-releases/statementsfrom-the-third-bipartisan-senate-forum-on-artificial-intelligence. UNI Europa. "AI Act - Our Submission to the EU Commission." June 30, 2021. https://www.unieuropa.org/news/ai-act-our-submission-to-the-eu-commission/. UNI Global Union. "Algorithmic Management: Opportunities for Collective Action." 2023. https:// uniglobalunion.org/wp-content/uploads/Algorithmic-Management-Opportunities-for-CollectiveAction.pdf. UNI Global Union. "First Collective Agreement on the Use of AI in Film and TV Production in Germany." March 14, 2025. https://uniglobalunion.org/news/first-collective-agreement-on-theuse-of-ai-in-film-and-tv-production-in-germany/. UNI Global Union. "H&M Workers Protected Under First Digitalization Agreement with ver.di." 2022. https://uniglobalunion.org/news/hm-workers-protected-under-first-digitalization-agreementwith-ver-di/. UNITE HERE Local 226. "Culinary Union Celebrates Historic Wins for Workers in the Best Contract Ever Won with MGM Resorts, Caesars Entertainment, and Wynn Resorts." Press release, November 10, 2023. https://culinaryunion226.org/news/press/culinary-union-celebrates-historic-wins-forworkers-in-the-best-contract-ever-won-with-mgm-resorts-caesars-entertainment-and-wynnresorts. Vallas, Steven P., Hannah Johnston, and Yana Moammadova. "Prime Suspect: Mechanisms of Labor Control at Amazon's Warehouses." Work and Occupations 49, no. 4 (2022): 421-456. van Kote, Gilles. "Le Monde Adopts a Charter on Artificial Intelligence." Le Monde, March 13, 2024. https://www.lemonde.fr/en/about-us/article/2024/03/13/le-monde-adopts-a-charter-onartificial-intelligence_6615286_115.html. Ver.di. "Erster Tarifabschluss zum Umgang mit KI in der Filmund Fernsehproduction." February 10, 2025. https://filmunion.verdi.de/und-action/nachrichten/++co++0a0e1f88-e78f-11ef-81ee-f5ae95fe3e8f. 85 ILO Working Paper 144 Vicioso, Dolores. "First Artificial Intelligence Bill Presented in Congress." dr1, February 20, 2025. https://dr1.com/news/2025/02/20/first-artificial-intelligence-bill-presented-in-congress/. Vincent, Catherine. "France: The Rush Towards Prioritising the Enterprise Level." In Collective Bargaining in Europe: Towards an Endgame, edited by Torsten Müller and Jeremy Waddington, 217-238. Brussels: European Trade Union Institute. All India IT & ITeS Employees’ Union (AIITEU). “BPO Workers’ Demand Charter.” Tech People, December 25, 2021. Voss, Eckhard, and Daniel Bertossa. 2022. "Collective Bargaining and Digitalization: A Global Survey of Union Use of Collective Bargaining to Increase Worker Control over Digitalization." New England Journal of Public Policy 34, no. 1: 10. Washington State Legislature. “House Bill 1672: Addressing Technology Used by Employers in the Workplace.” 2025–26 Regular Session. Introduced January 28, 2025. https://app.leg.wa.gov/ billsummary?BillNumber=1672&Year=2025&Initiative=false. Webster, Fiona, and Stephanie Rosseau. "Dominican Republic: Resolution Defines Teleworking Measures." Mercer. 2021, https://www.mercer.com/insights/law-and-policy/dominican-republicresolution-defines-teleworking-measures/. Weil, David. The Fissured Workplace: Why Work Became So Bad for So Many and What Can Be Done to Improve It. Cambridge, MA: Harvard University Press, 2014. Welsh Government. "Managing Technology That Manages People: A Social Partnership Approach to Algorithmic Management Systems in the Welsh Public Sector." 2024. https://www.gov.wales/ managing-technology-manages-people. White & Case LLP. "AI Watch: Global Regulatory Tracker - South Korea." June 18, 2024. https:// www.whitecase.com/insight-our-thinking/ai-watch-global-regulatory-tracker-south-korea. Wood, Alex J. "Algorithmic Management Consequences for Work Organisation and Working Conditions." JRC Working Papers Series on Labour, Education and Technology, No. 2021/07, 2021. Workforce Partnership Council. "Workforce Partnership Council Agreement: Partnership and Managing Change." December 2, 2021. https://www.gov.wales/workforce-partnership-councilagreement-partnership-and-managing-change-html. Wotschack, Philip, Florian Butollo, and Leon Hellbach. "Algorithmic Management and Democracy at Work in Germany." Incoding Policy Brief, 2024. https://ddd.uab.cat/record/290693. Writers Guild of America. "Summary of the 2023 WGA MBA." 2023. https://www.wga.org/contracts/ contracts/mba/summary-of-the-2023-wga-mba. Writers Guild of America East. "WGA East Members at FT Specialist Ratify First Union Contract." August 23, 2023. https://www.wgaeast.org/wga-east-members-at-ft-specialist-ratify-first-unioncontract/. Writers Guild of Canada. "Independent Production Agreement (IPA) 2024-2027, Appendix T." 2024. https://www.wgc.ca/sites/default/files/2025-01/2024%20IPA%20Final%20WGC_CMPA.pdf.