Developing countries' business participation in the AI economy

Lippoldt, Douglas Working Paper Developing countries' business participation in the AI economy CIGI Papers, No. 323 Provided in Cooperation with: Centre for International Governance Innovation (CIGI), Waterloo, Ontario Suggested Citation: Lippoldt, Douglas (2025) : Developing countries' business participation in the AI economy, CIGI Papers, No. 323, Centre for International Governance Innovation (CIGI), Waterloo (Ontario) This Version is available at: https://hdl.handle.net/10419/322457 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/ CIGI Paper No. 323 — June 2025 Developing Countries’ Business Participation in the AI Economy Douglas Lippoldt CIGI Paper No. 323 — June 2025 Developing Countries’ Business Participation in the AI Economy Douglas Lippoldt About CIGI The Centre for International Governance Innovation (CIGI) is an independent, non-partisan think tank whose peer-reviewed research and trusted analysis influence policy makers to innovate. Our global network of multidisciplinary researchers and strategic partnerships provide policy solutions for the digital era with one goal: to improve people’s lives everywhere. Headquartered in Waterloo, Canada, CIGI has received support from the Government of Canada, the Government of Ontario and founder Jim Balsillie. À propos du CIGI Le Centre pour l’innovation dans la gouvernance internationale (CIGI) est un groupe de réflexion indépendant et non partisan dont les recherches évaluées par des pairs et les analyses fiables incitent les décideurs à innover. Grâce à son réseau mondial de chercheurs pluridisciplinaires et de partenariats stratégiques, le CIGI offre des solutions politiques adaptées à l’ère numérique dans le seul but d’améliorer la vie des gens du monde entier. Le CIGI, dont le siège se trouve à Waterloo, au Canada, bénéficie du soutien du gouvernement du Canada, du gouvernement de l’Ontario et de son fondateur, Jim Balsillie. Credits Research Director, Digital Economy S. Yash Kalash Director, Program Management Dianna English Program Manager Grace Wright Manager, Publications Jennifer Goyder Publications Editor Christine Robertson Publications Editor Susan Bubak Graphic Designer Sami Chouhdary Copyright © 2025 by the Centre for International Governance Innovation The opinions expressed in this publication are those of the author and do not necessarily reflect the views of the Centre for International Governance Innovation or its Board of Directors. For publications enquiries, please contact [email protected]. The text of this work is licensed under CC BY 4.0. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. For reuse or distribution, please include this copyright notice. This work may contain content (including but not limited to graphics, charts and photographs) used or reproduced under licence or with permission from third parties. Permission to reproduce this content must be obtained from third parties directly. Centre for International Governance Innovation and CIGI are registered trademarks. 67 Erb Street West Waterloo, ON, Canada N2L 6C2 www.cigionline.org Table of Contents vi About the Author vi Acronyms and Abbreviations 1 Executive Summary 1 Introduction 5 Literature Review 10 AI Start-Ups and the Case Study Countries 32 Policy Assessment 34 Conclusions 38 Appendix 1: Key Sources 40 Appendix 2: Data 46 Appendix 3: OECD Recommendation of the Council on Artificial Intelligence 47 Works Cited vi CIGI Paper No. 323 — June 2025 • Douglas Lippoldt About the Author Douglas (Doug) Lippoldt is a CIGI senior fellow and an international trade economist based in Claremont, California. He served as chief trade economist at HSBC Global Research from 2014 to 2020. Previously, he served in various roles as a senior economist at the Organisation for Economic Co-operation and Development in Paris, France, during a tenure of 22 years. Doug’s early career included seven years as an international economist with the US Department of Labor. He has published extensively on trade topics as well as on related aspects of economic development, labour market adjustment, innovation and intellectual property. Doug holds a Ph.D. in economics from the Institut d’études politiques de Paris (Sciences Po), an M.A. in international studies from the University of Denver and a B.A. in international studies from Washington College in Maryland. He was a Fulbright Scholar at the University of Cologne, Germany, and a Peace Corps volunteer in Burkina Faso. During 2015–2019, he represented HSBC as the deputy delegate on the Business 20 Trade and Investment Task Force. He is currently a contributing author on a Think20 trade task force team. Acronyms and Abbreviations AI artificial intelligence ASEAN Association of Southeast Asian Nations BRI Belt and Road Initiative DEFA Digital Economy Framework Agreement DEPA Digital Economy Partnership Agreement DPA Digital Policy Alert ECLAC Economic Commission for Latin America and the Caribbean G7 Group of Seven GenAI generative AI IoT Internet of Things IP intellectual property IPO initial public offering LLM large language model MENA Middle East and North Africa ML machine learning NCAIR National Center for AI and Robotics NLP natural language processing OECD Organisation for Economic Co-operation and Development PPP purchasing power parity R&D research and development SMEs small and medium-sized enterprises STEM science, technology, engineering and mathematics UAE United Arab Emirates VC venture capital WEF World Economic Forum 1Developing Countries’ Business Participation in the AI Economy Executive Summary How are start-up businesses in middle-income developing countries participating in the rapidly expanding artificial intelligence (AI) economy? Are some developing country AI start-ups able to tap into the AI ecosystems of leading economies in Group of Seven (G7) countries?1 What is the domestic policy context, and how can conditions be improved? Are there signs of South-South AI ecosystem development? This case study takes stock of the development of AI start-ups engaged with G7 investors and investors from across 10 middle-income developing countries. It considers the emerging investment AI ecosystems in these countries, which are located across Africa, Southeast Asia and Latin America. The significant concentration of advanced AI innovation in developed economies heightens the importance of engagement with counterparts in G7 countries who can complement local resources with access to additional financial, technological and managerial support. The research employs a comparative case study approach analyzing firm-level data from 2,537 AI start-ups across Brazil, Colombia, Egypt, Indonesia, Kenya, Nigeria, South Africa, Thailand, Tunisia and Vietnam. Particular attention is given to the role of G7 investors in facilitating technology transfer, with an examination of 239 AI start-ups that have attracted such investment. The analysis is complemented by an assessment of AI policy and regulatory frameworks in each study country. From national strategies to implementation and enforcement, various dimensions are reviewed. Key findings indicate that: → Despite modest scale, viable AI ecosystems are developing in all the case study countries; South-South investment networks are emerging alongside dominant North-South relationships; and policy and regulatory development is broadly supportive. → Among the 10 countries, Brazil has achieved some scale; domestic investors there supply a 1 The G7 countries are Canada, France, Germany, Italy, Japan, the United Kingdom and the United States. majority of investment engagements2 with local AI start-ups, which is unique for these markets. Brazil attracts by far the largest inflow of US investor engagements, while in most of the other markets, US investors play a lead role. → There is some geographic clustering of these AI start-ups, pointing to the possible emergence of innovation hubs in and around cities such as Jakarta, Lagos and São Paulo. → All of the host countries are developing national strategies. Implementation progress varies widely. In all of the case study countries, challenges remain in relation to issues such as infrastructure constraints, skills development, the urban-rural digital divide, regulating AI safety while facilitating innovation, and capacity limitations of the public administration, among other issues. The analysis underscores that while these developing countries are pursuing diverse strategies for AI development, they are also facing challenges in striving for regulatory convergence with major AI powers, which are themselves diverging on some issues such as risk management. To succeed, the case study countries must balance international alignment with domestic innovation support, address infrastructure and talent shortages, and develop governance frameworks that protect citizens while enabling competitiveness in global AI markets. The study concludes with corresponding recommendations. Introduction The AI economy is advancing rapidly technologically and attracting substantial investment inflows. This is being fuelled, in part, by private sector research and development (R&D) expenditure in areas ranging from chipmaking equipment to software development and commercial applications (Lippoldt 2024). There is significant geographic concentration in large-scale AI innovative activity, which is centred in roughly a dozen mostly developed countries, together with 2 An engagement is defined here as a level of investment sufficient to place an investor among the top five in a firm. 2CIGI Paper No. 323 — June 2025 • Douglas Lippoldt China and India.3 Given the assessed potential for AI to contribute substantially to future economic growth, this concentration in AI innovation presents a challenge for developing countries to gain access to and share in the benefits of this innovation while also managing the risks. This geographic concentration could also pose a challenge for firms in other parts of the world looking to tap into the latest generation of AI technology. A question remains as to whether there is potential for improved policy alignment or other adjustment to facilitate further development and technology transfer in the sector, with appropriate safeguards. In this paper, using a comparative case study approach, the author examines how start-ups in a sample of 10 middle-income developing countries engage with the global AI economy and considers domestic policy conditions that facilitate their participation. These issues are of importance not only for the direct stakeholders in such businesses but also for those concerned with economic development more generally. International diffusion and application of the emerging AI-driven technologies from leading AI economies, along with complementary local innovation in other economies, may together contribute to improved welfare in developing countries, provided risks are managed. Such risks may come from intentional or accidental harms directly associated with AI (for example, due to biases), as well as other challenges or constraints such as regulatory impediments, adjustment costs for affected individuals and social resistance to technological change, among other considerations.4 Where to Plug In? Development of AI foundation models requires significant scale and a substantial resource base to amass the necessary computing power and 3 Lippoldt (2024, 14, 18) found the leading AI-intensive, innovation-driven firms (in terms of R&D expenditure) to be concentrated geographically. As of 2021, the corporate headquarters were located in just 11 countries: Canada, China, Finland, Germany, India, Ireland, Israel, Japan, South Korea, the United Kingdom and the United States. These firms were often clustered in particular regions, such as Beijing, London or Silicon Valley. 4 Ian Bremmer and Mustafa Suleyman (2023), for example, highlight key risks. data.5 Few firms or institutions in the world have sufficient capacity to tackle such a challenge. For example, according to Stanford University’s 2024 Artificial Intelligence Index Report (Maslej et al. 2024, 5), OpenAI’s GPT-4 required an estimated US$78 million worth of computational power to train in 2023. Even though some less expensive strategies for training appear to be emerging via challengers such as DeepSeek, the upfront costs remain substantial.6 In 2023, China, the European Union and the United States together produced less than 100 notable AI models, according to the 2024 AI Index Report (ibid.). But, fortunately, this is not the full story. There is room in the AI economy for smaller developers to build downstream applications that employ the larger models as a base.7 Businesses may also develop “small language models” that have a compact design and possess far fewer parameters than a large language model (LLM) — perhaps just one-sixth or less of the number of parameters. Nevertheless, they can excel at delivering AI services for specific tasks, such as analyzing customer feedback or generating product descriptions. And there are opportunities to be found in many similar niches in the delivery of AI services to customers in developing countries, from building out hardware and software capacity to human capital development (in other words, skills) and public infrastructure for connectivity, among other possibilities. Policy makers in developing countries have a substantial role to play in shaping the domestic conditions for business. Their choices may contribute to a nation’s ability to tap into the benefits of the AI economy while safeguarding against risks. Of course, AI is not a stand-alone solution to the full range of constraints on 5 In a web post titled “Reflections on Foundation Models,” Rishi Bommasani and Percy Liang (2021) of Stanford University’s Institute for HumanCentered Artificial Intelligence define foundation models as “models trained on broad data (generally using self-supervision at scale) that can be adapted to a wide range of downstream tasks.” They “see foundation models as the subject of a growing paradigm shift, where many AI systems across domains will directly build upon or heavily integrate foundation models.” 6 Some foundation model developers have devised strategies to reduce training costs, in part, by trading off some refinement in the results. Chinese start-up DeepSeek famously trained its debut model for just US$6 million, though that number allegedly excludes substantial upfront costs incurred prior to the launch of training. See Metz (2025). 7 For example, see the discussion in a web post from Amazon Web Services (https://aws.amazon.com/what-is/foundation-models/) and from the World Economic Forum (Whiting 2025). 9Developing Countries’ Business Participation in the AI Economy regulation that supports both export-oriented and domestically oriented industrialization. Ezekiel T. Mutasa, Chitra Dhiwwale and Sundaran Sagaran A. Gopal (2024) examine AI applications, prospects and challenges in developing economies across key sectors, including manufacturing, agriculture, retail, financial services, health care and mining. They take a meta approach, drawing on empirical studies that mostly date from the period prior to the emergence of publicly accessible GenAI. The authors identify significant benefits of firmlevel AI adoption, such as increased productivity and innovation. The net gains are derived via a variety of channels such as improved decision making, quality control, predictive maintenance, tools for optimization of resource use and to reduce wastage, better targeting of underserved populations and improved health-care screening. They also highlight substantial challenges to successful firm-level AI adoption, including data privacy concerns, high implementation costs, inadequate digital infrastructure, skill shortages and resistance to change. Mutasa, Dhiwwale and Gopal (2024) highlight policy recommendations for middle-income countries: → encourage AI adoption while addressing ethical and sociocultural considerations; → invest in critical infrastructure, especially in rural areas; → employ change management strategies to address resistance to adoption; → support training programs to bridge the AI talent gap; → use public-private partnerships to mitigate high costs; and → address cross-sectoral challenges such as poor data quality and integration. At the firm level, they advise that management should: → look for sector-specific opportunities to apply AI; → build on existing infrastructure (for example, in mobile technology); → focus on practical, high-impact solutions; → tailor solutions to local contexts (for example, underserved domestic populations); and → seek collaboration between government, private sector and other local stakeholders to overcome adoption barriers and scale AI solutions. Xueyuan Gao and Hua Feng (2023) consider the rollout of AI in China during the period prior to the recent leap forward in GenAI capabilities. They examine how AI adoption affects manufacturing firms’ productivity in China. Using microlevel data from 2010 to 2021, they found that each percentage increase in AI penetration was associated with large gains in total factor productivity. The authors identified three mechanisms for this: value-added enhancement (improving product quality and production processes); skill-biased enhancement (shifting toward higher-skilled workers); and technology upgrading. Implementation of AI was found to stimulate innovation, including with respect to further AI innovations. Regarding AI policy, the authors made the following recommendations: → target any AI subsidies at capitaland technology-intensive industries, which benefit more than labour-intensive ones; → consider market structure (firms in industries with high concentration experienced greater productivity gains from AI, perhaps due to scale effects); → take firm ownership into account, as private enterprises realized significant productivity improvements while state-owned enterprises did not, suggesting institutional reforms may be needed alongside technology adoption; and → accompany AI implementation with complementary investments in human capital (AI drives demand for highly skilled workers). David Heller and Dominik Asam (2024) examine the impact of GenAI on start-up productivity, comparing firms in the software sector to other sectors. Using GitHub Copilot’s release as a quasi-natural experiment, the authors employed Crunchbase data covering 21,834 start-ups that secured initial funding between Q1 2020 and Q3 2023. They found that software-developing start-ups experienced a 20 percent reduction in time-to-funding (an early indicator of productivity) relative to other start-ups prior to Copilot’s release. The effects were most pronounced for 10 CIGI Paper No. 323 — June 2025 • Douglas Lippoldt start-ups with founders possessing technological or managerial experience, suggesting that GenAI serves as a competitive advantage when combined with complementary human capital. Their findings indicate that GitHub Copilot may substitute for traditional resources such as junior programmers while enhancing the productivity of experienced entrepreneurs. This is of particular importance during early start-up stages when resource constraints are most binding. Nicholas Otis et al. (2024) conducted a field experiment with 640 Kenyan entrepreneurs to evaluate the impact of a GenAI business assistant (GPT-4-powered) delivered via WhatsApp. While the study found no significant average treatment effect on business performance (measured as profits and revenues), it revealed substantial heterogeneity based on pretreatment performance levels. High-performing entrepreneurs tended to benefit overall, with an approximately 15 percent improvement in performance. Low performers tended to experience negative performance, with an overall decline of eight percent. The divergent effects stemmed from the manner in which the entrepreneurs selected and implemented the AI’s suggestions. High performers tended to better identify opportunities for specific improvements, while low performers disproportionately implemented generic advice (for example, price discounts) that harmed their businesses. Literature Review: A Summing Up This review of the literature spans economic impacts, governance challenges and implementation hurdles. It provides a basis to consider the transformative opportunities opened by AI for middle-income countries (for example, improved productivity, innovation and sector-specific solutions), as well as the risks (for example, widening inequality, job displacement and regulatory challenges). Tension between AI nationalism and international cooperation is highlighted, with scholars such as Aaronson noting how divergent regulatory approaches may create market access barriers for developing nations. Firm-level assessments reveal that successful AI adoption depends, in part, on contextual factors including skills development, infrastructure readiness and appropriate regulatory frameworks. Several studies emphasize that while AI has the potential to accelerate economic development, this outcome requires deliberate policy design balancing innovation with ethical governance (especially regarding data governance challenges). The link between AI start-up development and national economic welfare can be seen via research showing that firms with local knowledge can create AI solutions tailored to local contexts. This may potentially create competitive advantages for such firms while also addressing national development priorities such as health care, agriculture and financial inclusion (for example, Mayer 2021; Mutasa, Dhiwwale and Gopal 2024; Mannuru et al. 2023). AI Start-Ups and the Case Study Countries This section presents the firm-level analysis for the case study. It begins with the principal exercise: a review of AI start-up firms with G7 investor engagement in the case study countries. This is followed by a brief assessment of AI unicorn firms, with consideration of these successful firms in relation to the case study of AI ecosystems. Case Study Countries The central focus here is on the prospects for developing countries to connect and integrate into the rapidly expanding AI economy, taking into account the role of AI start-ups. These are young, active, privately held businesses. They constitute a category that has demonstrated a particular dynamism in the field of AI. They are also likely to play an important role in the integration of AI into developing economies. For example, this might arise via the exploitation of niche opportunities to develop tailored AI applications or small language models to address local conditions in a developing country or region.21 Given that many high-income countries are already leading in AI developments or positioning for wide adoption of AI, this case study targets the next tiers of countries by income. In order to ensure diversity of coverage, an illustrative sample of 10 middle-income countries was selected from across Asia, Africa ((the Middle East and North Africa [MENA] and Sub-Saharan 21 The value of exploiting this type of opportunity is supported by some of the empirical work cited in the literature review above (for example, Mayer 2021; Mutasa, Dhiwwale and Gopal 2024). 11Developing Countries’ Business Participation in the AI Economy Africa) and Latin America. The sample includes five lower-middle-income countries and five uppermiddle-income countries, as classified by the World Bank (Table 1). Box 2 presents an overview of the countries not selected for inclusion in the sample. The author’s selection of specific countries was further guided by data availability with respect to start-up firms in the AI sector and the availability of standardized AI policy and regulatory information. Data for the firm-level analysis was drawn from the Crunchbase data set,22 which covers start-ups globally and has detailed descriptive information about start-up firms and their investor counterparts. In order to target markets with at least nascent AI ecosystems, the author selected sample countries with more than 50 start-up firms listed as having the key words “artificial intelligence” in their mission or purpose descriptions. A complementary start-up analysis was developed, drawing on the global unicorn data set produced by CB Insights.23 This covers 22 For a brief overview of Crunchbase data strengths and weaknesses, see Appendix 1. 23 See www.cbinsights.com/research-unicorn-companies. start-ups that are still private and that have achieved a market valuation of US$1 billion or more. Finally, with respect to policy analysis, the research turned primarily to the OECD AI Policy Observatory, which has broad country coverage, and standardized and readily accessible policy and regulatory data.24 This was supplemented and updated using the Digital Policy Alert database25, as well as regional and national sources. On this basis, the country sample was established to include Brazil and Colombia; Egypt, Kenya, Nigeria, South Africa and Tunisia; and Indonesia, Thailand and Vietnam (Figure 1). As can be seen in Table 1, these nations represent a broad range in terms of economic scale and population. The sample is illustrative of a diverse group of countries that have — to varying degrees — achieved some 24 For an overview of the content and country coverage of the OECD AI Policy Observatory database, see Appendix 1 and https://oecd.ai/en/dashboards/overview. 25 For an overview and a link to the database website, see Appendix 1. 26 The economic data cited in this paragraph is drawn from the author’s tabulations and the World Bank’s World Development Indicators, available at https://data.worldbank.org/indicator. The AI start-up data is from Crunchbase and is available at www.crunchbase.com/. 27 The numbers in parentheses represent the number of AI start-ups in each of these countries as of Q1 2025. Box 2: Economies Not Selected for the AI Start-Up Sample The 91 middle-income developing countries not covered directly in this case study represent a slice of the global economy that is roughly five times greater than the collective share of the case study economies (US$30.6 trillion versus US$6.1 trillion in 2023, in current US dollars).27 The middle-income countries not covered were somewhat less well off, on average, than the case study countries by some indicators. For example, those not covered had a median GDP (purchasing power parity [PPP]) per capita of US$11,245 in 2023 versus US$15,304 for the sample countries. A portion of this differential may reflect a selection bias in that the author set out to find middle-income countries that had already demonstrated some engagement in the AI economy, with at least 50 or more AI start-ups. This choice was a natural consequence of the author’s research design to examine the nascent AI sector activity in an illustrative sample of middle-income countries. While some of those not selected were competing in the AI sector (for example, Pakistan with 262 AI start-ups, Ghana with 53 or Peru with 70), many countries not selected have each faced a unique combination of challenges that prevented the emergence of an adequate number of AI start-ups. Examples of middle-income developing countries with less than the author’s minimum selection threshold of 50 AI start-ups and with GDP (PPP) per capita below the author’s AI sample median include Honduras (two), Laos (zero), Libya (one) and Tanzania (10).28 Some challenges to AI start-up development — for example with respect to market openness or human capital development — may also have economic causes or consequences and that may be reflected in the lower median per capita incomes. 12 CIGI Paper No. 323 — June 2025 • Douglas Lippoldt traction in the AI economy. Their situation may offer some useful insights for further development of the AI economy in other developing nations. Context for AI Implementation in Developing Countries There is broad awareness of AI across a substantial share of the population in developing countries. This is evidenced, for example, by relatively frequent use of Chat GPT-4 (Maslej et al. 2024, chapter 9, 449). As can be seen in Table 2, for countries covered by the Ipsos survey data,28 a majority of adults feel positively about AI benefits and trustworthiness. This may be associated with the early stage of the sector’s development in these countries, but it certainly contrasts strikingly with the view of much of the public in the more advanced economies. 28 See Appendix 1 for details of the Ipsos survey. In our sample countries, some of the optimism may be associated with the youthfulness of the population. For most of these countries, the median age is around the global median or younger (Thailand is an exception in this regard).29 Age appears to be one factor influencing attitudes to AI, a point noted in Stanford University’s 2024 AI Index Report (ibid., 438). Indeed, it may be that such public awareness and positive attitudes contributes an impulse toward interest in AI entrepreneurship in the sample countries. These positive attitudes seem to have carried over to many current business leaders. A survey of business executives by the World Economic Forum (WEF) in 2024 found that respondents in eight of the sample countries did not rank “risk of adverse outcomes from AI technologies” as a 29 This is based on data available at www.cia.gov/the-world-factbook/field/ median-age/country-comparison/. Figure 1: Case Study Countries Colombia Brazil Egypt Indonesia Kenya Nigeria South Africa Thailand Tunisia Vietnam Source: Microsoft Excel map; author’s tabulations. 13Developing Countries’ Business Participation in the AI Economy Table 1: Case Study Countries, Overview Economy Income Status GDP Population AI Start-Up Firms, as Recorded in the Crunchbase Database FY2025, Based on 2023 per Capita GDP 2023, Current US$, Billions Percent of World Total 2023, Millions Percent of World Total Number Active, as of 1Q2025 Percent of World Total Brazil Upper middle income 2,173.7 2.0 211.1 2.6 973 1.1 Colombia Upper middle income 363.5 0.3 52.3 0.6 348 0.4 Egypt Lower middle income 396.0 0.4 114.5 1.4 98 0.1 Indonesia Upper middle income 1,371.2 1.3 281.2 3.5 199 0.2 Kenya Lower middle income 108.0 0.1 55.3 0.7 79 0.1 Nigeria Lower middle income 363.8 0.3 227.9 2.8 230 0.3 South Africa Upper middle income 380.7 0.4 63.2 0.8 294 0.3 Thailand Upper middle income 515.0 0.5 71.7 0.9 80 0.1 Tunisia Lower middle income 48.5 0.0 12.2 0.2 50 0.1 Vietnam Lower middle income 429.7 0.4 100.4 1.2 186 0.2 World total (all economies) 106,170.0 100.0 8,061.9 100.0 86,235 100.0 Source: www.crunchbase.com/discover/organization.companies; Metreau, Young and Eapen (2024); https://data.worldbank.org/indicator/NY.GDP.MKTP.CD2024; https://data.worldbank.org/indicator/SP.POP.TOTL. Note: For the fiscal year beginning July 1 2024, the World Bank defines lower-middle-income countries as having gross national income per capita ranging between US$1,146 and US$4,515; upper-middle-income countries are defined as having gross national income per capita ranging between US$4,516 and US$14,005. Tunisia’s share in world total GDP is 0.05. Crunchbase relies on a variety of sources and is in part crowdsourced, so there may be variation in the quality of the data from country to country (see Appendix 1). 14 CIGI Paper No. 323 — June 2025 • Douglas Lippoldt Table 2: Popular Perceptions of AI Public feelings about AI Products and services using AI have more benefits than drawbacks (% agree “very” or “somewhat”) Public trust in AI I trust companies that use AI as much as I trust other companies (% agree “very” or “somewhat”) Executives' opinions on "risk of adverse outcomes of AI technologies": Is this a top-five risk over the next two years in your country? If “yes,” then rank is given (1 = most cited to 5 = fifth most cited); if “no,” then “no” Global country (simple) average 54 52 Yes, #1 Indonesia 78 69 Yes, #1 Thailand 74 73 No Mexico 73 66 No Malaysia 69 70 No Peru 67 60 No Türkiye 67 65 No South Korea 66 55 No Colombia 65 56 No India 65 67 No Brazil 64 60 No Singapore 64 57 Yes, #5 Romania 61 62 No South Africa 59 55 No Chile 59 51 No Argentina 57 52 No Italy 55 53 No Japan 52 44 No Spain 50 49 No Hungary 48 46 No Poland 47 50 No Great Britain 46 45 Yes, #4 New Zealand 44 43 No The Netherlands 43 44 No Germany 42 45 No Ireland 40 39 No Australia 40 42 No Belgium 39 39 No Sweden 39 42 n/a Canada 38 39 Yes, #5 France 37 37 No United States 37 36 Yes, #3 Egypt n/a n/a No Kenya n/a n/a No Nigeria n/a n/a No Tunisia n/a n/a No Vietnam n/a n/a Yes, #1 Source: Public feelings and public trust: Ipsos (2023, 9 and 15); Executives’ opinion of AI risk: Elsner et al. (2025, Appendix C, 81–91). Note: See Appendix 1 for details of the Ipsos survey and the WEF survey. 15Developing Countries’ Business Participation in the AI Economy top risk over the next two years (Table 2).30 The exceptions were Indonesia and Vietnam, where executives did rate AI as the top risk. This may be associated with the development and advancing implementation of AI governance in the particular cultural context of those countries which may raise awareness of risks, as well as sector-specific risk issues, among other possible explanations. Obstacles to Business Start-ups are young companies and vulnerable to a variety of challenges, such as exhausting their liquidity, misjudging market demand for a product and unfavourable regulatory changes, among many others. Start-ups often fail. Even in countries such as the United Kingdom and the United States, many start-ups — and, in some years, most — do not survive past their fifth or sixth anniversaries.31 In the study countries, it is unlikely that performance is much better. To provide a glimpse into business perceptions of obstacles in the case study countries, the paper turns to the World Bank Enterprise Surveys. These have been conducted in dozens of countries, including each of the 10 case study countries and three of the G7 economies. China is included here for the sake of comparison. Table 3 presents a tabulation of firms’ perceptions in each country of their biggest obstacles. Businesses are surveyed across a broad range of sectors and firm sizes and do not necessarily have a focus on AI (see the table notes for details). For most of the countries, the obstacles shown account for a majority of the top concerns cited by businesses. The table also reveals a striking contrast between the case study countries and the G7 countries. For most of these nations, except Brazil, “political instability,” “access to electricity” and “access to finance” are among the leading categories of obstacles among those 30 Respondents could select risks from among 34 options across five categories including economic, environmental, geopolitical, societal and technological risks. See the Key Sources Annex for details of the WEF survey. 31 For example, among all US private businesses established in in the year ending March 2018, only 51.9 percent survived five years on, and by March 2024, only 47.5 percent survived. See www.bls.gov/bdm/ us_age_naics_00_table7.txt, accessed October 30, 2024. According to the World Bank Group “Prosperity Data360” online database, the five-year survival rate for UK businesses in the “total industry, construction and market services except holding companies” sector was just 31 percent as of 2018 (in other words, 69 percent had failed). See https:// prosperitydata360.worldbank.org/en/indicator/OECD+BDI+YS5_R. shown in the table. One or more of these basic operational considerations was cited by more than 15 percent of businesses surveyed in each of the case study countries (excluding Brazil). In Egypt, Kenya and Vietnam, informal sector competition also weighed as a top concern, being cited by more than 15 percent of businesses in those three nations. Chinese businesses also cited access to finance and informal sector competition as top concerns. In addition, in three of these economies, 10 percent or more of businesses ranked “corruption” as the top obstacle. Brazil was an outlier among the developing nations, with a profile more closely resembling the G7 nations shown. For the G7 economies covered in the table, those issues were much less frequently cited by respondents. Instead, many G7 firms pointed to shortfalls in the availability of adequately trained workforce participants as being a top concern, by far. It is notable, however, that more than 10 percent of businesses in Brazil, China and Vietnam also cited this as a top concern. This latter point is an oft-cited concern for firms developing or implementing AI systems around the world.32 AI start-ups may be particularly vulnerable to constraints such as high capital needs, patchy broadband and limited availability of human resources, which can directly hamper AI solution deployment or scaling. For example, a recent Brookings Institution study on leveraging AI to support Africa’s economic development (Signé 2025) points to the following constraints. → limited digitized data availability with respect to Africa (for example, only 0.02 percent of total internet content is in African languages); → limited availability of digital skills and relevant human capital (universities in Africa are introducing AI courses, but often there is a lack opportunities for hands-on learning); and → limited R&D expenditure flows to Africa, with African use-case development often neglected. Moreover, such challenges are magnified by operational constraints related to poor infrastructure and high costs to address the bottlenecks (for example, to develop cloud computing capacity). Start-ups seeking to address 32 For more on AI sector skills demand, see Maslej et al. (2024, chapter 4, section 2, on jobs; chapter 6 on education) 16 CIGI Paper No. 323 — June 2025 • Douglas Lippoldt these challenges may be constrained by the very conditions in which they are operating, which contribute to risk aversion on the part of some investors and financial institutions, and shortfalls in the availability of needed capital. Starting with a Few Definitions Before proceeding to the detailed analysis, the paper first presents a few definitions. For the purposes of this analysis, the paper defines “AI start-up” as a firm that has “AI” integrated as part of its mission statement or purpose. The paper did not impose a strict age limit, but in the study population, the median AI start-up age by country is young, ranging from five years in Colombia to nine years in South Africa. The paper only considered firms that are currently active. Generally, the AI start-up firms in the assessment were not at the more mature stages of commercial development, such as preparing for an initial public offering (IPO) (see Box 3). Unicorns are considered in a subsequent section and tend to be more mature, in the mid-tolate stages of development for a start-up. Table 3: Business Perceptions of Their Biggest Obstacles, Selected Concerns, Most Recent Year Available (% of Firms Responding) Country Access to Electricity Corruption Access to Finance Inadequately Educated Workforce Informal Sector Political Instability Subtotals (out of 100%, by Country) Brazil 0.3 3.3 7.5 12.6 12.4 3.0 39.1 Colombia 4.6 8.7 7.5 5.2 8.2 39.0 73.2 Egypt 2.9 6.2 10.7 2.2 15.5 25.7 63.2 Indonesia 1.6 10.1 28.8 3.4 7.3 11.3 62.5 Kenya 3.0 7.7 18.3 1.6 22.9 17.0 70.5 Nigeria 27.2 12.7 30.2 0.4 4.3 4.4 79.2 South Africa 54.6 5.8 16.1 0.0 0.8 13.3 90.6 Thailand 19.8 2.3 4.6 2.6 5.0 20.3 54.6 Tunisia 0.6 15.0 39.4 5.4 8.3 11.5 80.2 Vietnam 4.2 0.3 21.2 11.7 22.1 3.6 63.1 Dev’g Country Avg. 11.9 7.2 18.4 4.5 10.7 14.9 55.7 China 4.8 1.2 22.4 13.0 19.6 0.8 61.8 France 3.3 1.5 2.8 23.8 11.1 4.3 46.8 Germany 0.2 0.4 5.0 53.5 4.1 6.2 69.4 Italy 4.5 1.1 5.4 18.8 3.8 5.5 39.1 Source: Source: World Bank (2025), “Enterprise Surveys,” Global Indicators Department, Data Visualization, https://www. enterprisesurveys.org/en/graphing-tool. Notes: 1) In the Table, “Dev’g Country Avg” refers to the average scores across the 10 case study countries. “Subtotals” refers to the tally of percentages for the six categories shown for each country. For most countries, the table captures a majority of the top obstacles identified by respondents (NB, each respondent could only select one top obstacle). 2) The surveys cover registered firms with five or more employees and one percent or more of private ownership. The survey covers most of the private sector. See Appendix 1 for details. 3) The survey respondents could select among 15 categories of biggest obstacle. The six presented here were selected for their generally high frequency in the case study countries and their relevance to AI firms (e.g., as opposed to items such as “access to land” or “crime, theft and disorder”). See the Data Visualization page, linked above, for the full list of obstacles for which data is available. By subtracting the subtotal number from 100 percent for each country, the reader can obtain the total value of the omitted categories. 4) Survey years are as follows: Brazil (2009), China (2012), Colombia (2023), Egypt (2020), France (2021), Germany (2021), Indonesia (2023), Italy (2024), Kenya (2018), Nigeria (2014), South Africa (2020), Thailand (2016), Tunisia (2020), Vietnam (2023). 17Developing Countries’ Business Participation in the AI Economy In looking at the commercial linkages of the AI start-ups, the paper focuses on “investor engagement,” defined as a case where an investor has sufficiently supported an AI start-up to reach the level of being a top-five investor in the firm. This may involve one or more “deals,” which refer to transactions where an investor provides capital in exchange for an equity stake in the firm or convertible debt. After a general introduction to capital flows in the sector (including deals), the paper then turns to use investor engagement for the assessment of the AI start-ups. The author chose investor engagement rather than deal counts in order to focus on the more substantial financial relationships, which can entail additional commercial support for an AI start-up. Figure 2: Total Quarterly AI Funding Deal Counts, Q1 2020–Q3 2024, Shares by Region (%) 0 10 20 30 40 50 60 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 2020 2021 2022 2023 2024 Asia Europe Canada United States Latin America Canada All Other RegionsAfrica Europe AsiaUnited States Sources: See www.cbinsights.com/reports/CB-Insights_AI-Report-2024.xlsx; author’s tabulations. Note: The Africa series is derived using proportions from the deal data for 2020 to 2024 in CB Insights Q3 2024 applied to the updated data set in www.cbinsights.com/reports/CB-Insights_AI-Report-2024.xlsx. 18 CIGI Paper No. 323 — June 2025 • Douglas Lippoldt Situating Our Case Study in the Global VC Context In order to situate the assessment of AI start-ups in the 10 case study countries, it is useful to start with a global perspective. Figure 2 presents an assessment of the global shares in AI start-up deal funding by region. As can be seen, the United States has a disproportionate share of the total, followed by Asia and Europe. Canada is next, followed by Latin America, all other regions (including Oceania) and Africa. The exact ranking depends on developments in each quarter. But it can be seen from the data that two of the three regions home to the study countries account for just a small share of the global VC flowing to AI start-ups. (A separate breakdown from Asia for the third region, Southeast Asia, is not available in this data set.) The comparatively small numbers do not mean that the case study countries are unimportant for AI, however. In the first place, they collectively have some market scale, accounting for 1.2billion people and US$6.2 trillion of GDP (in other words, just under six percent of global GDP) (see Table 1). Moreover, as noted above, most of these countries have youthful population age profiles that fall around or below the global median age, meaning that these markets may be positioned to be quite dynamic going forward. (Thailand is the exception to this metric, with a median age 10 years above the global average.) In addition, there is some notable AI development underway already, including AI entrepreneurship and business linkages to advanced economy partners, as well as national-level AI policy actions. To give a sense of the scale of AI start-up activity, Table 4 provides an overview of AI start-ups and VC deals by region for two recent quarters (Q3 and Q4 2024).33 This analysis reveals that there is a certain amount of activity in Latin America and Africa hinting at potential interlinkages to the financial mainstream. Running out of cash and failure to raise new capital are the top reasons start-ups fail, accounting for nearly two-fifths of all failures (CB Insights 2021, 4). These funding deals play an important role for start-ups seeking to avoid such a fate. But they can also help an AI start-up to plug into the mainstream. They can deliver non-financial benefits in the form of advice related to technology, managerial skills, operations and network connections (for example, partners in new markets). A Look at Case Study AI Start-Up Profiles The assessment of case study country start-ups proceeds in two parts. The first section — the 33 It should be noted that there is a lot of quarter-to-quarter volatility in VC markets. These two recent quarters are presented here for illustrative purposes and not as an indication of larger trends. Table 4: AI-Related VC Funding and Deals, By Region, Q3 and Q4 2024 Q3 2024 Q4 2024 Deals (Counts) Funding (US$ millions) Shares of Total Funding (%) Deals (Counts) Funding (US$ millions) Shares of Total Funding (%) United States 566 11,442 68.0 548 38,028 86.9 Europe 279 2,779 16.5 291 2,519 5.8 Asia 316 2,112 12.5 270 2,045 4.7 Latin America 29 62 0.4 17 92 0.2 Canada 27 184 1.1 24 922 2.1 Africa 7 9 0.1 1 0 0.0 All other regions (including Oceania) 21 245 1.5 14 160 0.4 Totals 1,245 16,832 100.0 1,165 43,766 100.0 Source: www.cbinsights.com/reports/CB-Insights_AI-Report-2024.xlsx; author’s tabulations.. 25Developing Countries’ Business Participation in the AI Economy study countries. Altogether, these local investors contributed 26 investor engagements in other case study countries. While the numbers are modest, it is interesting to see the emergence of some SouthSouth investment flows in support of AI start-ups. There are the beginnings of regional ecosystems operating, especially among the five African case study countries. There are also African ties to Latin America with some investor engagement in both directions. This type of activity is lacking among the three Southeast Asian case study countries. To get a better handle on the overall flows, we now turn to examine the top sending countries for investment engagements received by each of the case study countries (see Table 9).38 The United States dominates the rankings for lead investor in seven countries. But three of the case study 38 Appendix 2 presents detailed profiles of the top G7 investors (Table A.1); the top domestic investor in each case study country, except Thailand whose covered firms only had foreign investors (Table A.2); and the top investors across all the sample countries (Table A.3). countries — Brazil, Indonesia and Tunisia — are their own lead investor, with the United States placing second. The strong local performance in these countries may provide an indication of a dynamic local ecosystem emerging for AI start-ups. By a wide margin, the strong United States-toBrazil channel accounts for the largest single international flow of investor engagements shown in the table. US investors also showed relatively strong interest in Nigeria and South Africa. Altogether, five G7 countries are represented across the table. Japan, in particular, is represented on investment flows to all three case study countries in Southeast Asia, indicating a potentially important regional corridor. (Some three quarters of Indonesia’s investor engagement originate in Asia.) There are further indications of South-South regional hub flows via investor engagements such as Saudi Arabian engagement in Egypt, Mauritian engagement in South Africa and Singaporean engagement in Thailand. Table 8: Domestic Investors Based in the Case Study Countries and Investing in G7 Affiliated AI Start-Ups at Home or Abroad in Other Case Study Countries, 1Q2025 Home Country Case Study Country Domestic Investor Entities (Count) Case Study Domestic Investors Investing Domestically (Count) Domestic Investment Engagements (Count) Outbound Foreign Investment Engagements by Domestic Firms to Other Study Countries (Count) Destination Countries Brazil 132 130 201 3 Kenya (1), South Africa (1), Tunisia (1) Colombia 3 3 3 0 Egypt 12 12 17 0 Indonesia 19 19 24 0 Kenya 7 4 4 4 Egypt (1), South Africa (1), Nigeria (2) Nigeria 15 13 16 5 Brazil (1), Colombia (2), Egypt (1), Kenya (1) South Africa 24 21 23 8 Kenya (6), Nigeria (2) Thailand 0 0 0 0 Tunisia 6 4 5 6 Egypt (5), South Africa (1) Vietnam 4 4 5 0 Total 222 210 298 26 Source: Crunchbase database; author’s tabulations and database construction; Claude.ai (assisted in the compilation of this table). Note: “G7 affiliated” means there is at least one investor from a G7 country investing in the Al start-up. The investor database assembled by the author to cover entities investing in Al start-ups in the 10 case study countries includes 526 unique investors. There were 210 domestic investors investing domestically, and there were 316 domestic and foreign investors investing in the 10 case study countries across international borders. 26 CIGI Paper No. 323 — June 2025 • Douglas Lippoldt Investment and Non-Financial Assistance The investor profile data reveals that investor engagements often involve assistance beyond purely financial investment. Many of these investors have gained experience internationally, which they are bringing to bear in support of their chosen AI start-ups. While some investors mention provision of specialist support for specific issues such as governance, many offerings fall into two broad clusters of issues: technology transfer and ecosystem development. Technology transfer takes a variety of forms. When large technology companies such as Google or Microsoft invest, they may provide access to their technology stacks and platforms.39 Only about three percent of the investors in our sample explicitly mention technology or skills transfer as a focus of their service. Most often this takes place through knowledge sharing, mentorship, and network 39 As noted above, AI start-ups covered in our study have generally not declared use of patent protection. In comparison, the international corporate investors are more likely to employ formal forms of intellectual property to protect their own interests. This means that the provision of adequate levels of intellectual property protection may be a useful part of the host country policy mix in order to facilitate access for domestic AI start-ups to technological inputs from abroad, along with the know-how on deployment (Park and Lippoldt 2005, 2014). connections rather than formal technology transfer programs. For the G7 investors, in particular, management expertise transfer is often a major component of their investment process. Among other elements, this may include strategic guidance and methodological support for operational scaling in relation to technological upgrades. Ecosystem development is another characterization that some investors give to their support. This means building mutually supportive relationships with the various counterparts that an AI start-up might need to engage in successfully pursuing its operations. Ecosystem development is an offering explicitly mentioned by 24 investors. Other dimensions of this type of assistance include network access (cited by 43 investors, this may include connections to potential customers, partners and follow-on investors), acceleration services (offered by 10 investors, which may include optimizing the implementation of AI technology), and incubation programs (offering suites of services using a holistic approach, as cited by five investors). The technology transfer and ecosystem development support cited by these investors in our case study firms aligns with the literature cited above on firm-level AI adoption in developing countries. As Mutasa et al. (2024) Table 9: Leading Investment Corridors (Counts of Investor Engagements in G7 Affiliated AI StartUps) Country Top Investor 2nd Investor 3rd Investor Top 3 Share (%) Brazil Brazil (200) United States (153) United Kingdom (8) 90.5 Colombia United States (10) Colombia (3) Spain (2) 75.0 Egypt United States (14) Egypt (13) Saudi Arabia (6) 60.0 Indonesia Indonesia (25) United States (12) Japan (9) 78.0 Kenya United States (17) South Africa (6) Kenya (4) 58.7 Nigeria United States (31) Nigeria (16) Canada (4) 76.1 South Africa United States (27) South Africa (22) Mauritius (4) 74.6 Thailand United States (2) Singapore (1) Japan (1) 100.0 Tunisia Tunisia (5) United States (4) France (2) 64.7 Vietnam United States (12) Japan (7) Vietnam (5) 75.0 Sources: Crunchbase, online database, www.crunchbase.com/discover/organization.companies (Q1 2025); author’s tabulations and investor database construction; Claude.ai provided final table compilation assistance. Note: Numbers in parenthesis refer to investor engagements. “G7 affiliated” means there is at least one investor from a G7 country investing in the Al start-up. 27Developing Countries’ Business Participation in the AI Economy highlight, effective AI implementation requires not just funding but also knowledge sharing, strategic guidance and network connections, among other elements. Certain of the investor offerings correspond to inputs that Fan and Qiang (2024) cite as necessary for promoting positive outcomes in the AI economy, including local ecosystem development and skills development. In addition, the emerging South-South investment observed in our case study countries also lends support to Mayer’s (2021) observation that firms with local knowledge can develop a measure of competitive advantage — and, potentially, improved resilience — by tailoring AI solutions to market conditions in developing countries. In sum, this complementary relationship between investors’ financial and non-financial inputs addresses a variety of needs at AI start-ups and potentially more generally in their home countries. Geography Matters Economist Paul Krugman and others have noted agglomeration effects, whereby development in a crowded technology centre may offer firms some advantages in terms of development and diffusion of innovative products (Krugman 1995). This likely also operates in the AI economy, as well. For example, thick labour markets may develop and offer large pools of sector-relevant talent. Improved communication around innovation could emerge due to the proximity of stakeholders, thereby conferring further information advantages. The geographic concentration may be supported by the availability of VC funding in some of these areas. The combination of ample investment funding and advisory support may provide a draw to specific geographies. Other draws might include the availability of access to complementary academic research. With respect to the case study sample of G7 investor-supported AI start-ups (see Table 10: Geographic Concentration of AI Start-Ups Country Total Start-Ups with G7 Investment City Number of Start-Ups, by City Market Shares (%) Brazil 119 São Paulo 75 63.0 Curitiba 9 7.6 Rio De Janeiro 4 3.4 Other cities 31 26.1 Colombia 9 Medellín 5 55.6 Bogotá 4 44.4 Egypt 15 Cairo 9 60.0 Alexandria 3 20.0 Gîza 2 13.3 Zamalek 1 6.7 Indonesia 18 Jakarta 9 50.0 Jakarta Pusat 2 11.1 Other cities 7 38.9 Kenya 14 Nairobi 14 100.0 28 CIGI Paper No. 323 — June 2025 • Douglas Lippoldt Table 10), one can observe some clustering. In eight of the 10countries, one city accounted for a majority of the activity. This was most pronounced in São Paulo, Brazil, with 75 AI start-ups as of 1Q2025. In South Africa and Vietnam, there was a more balanced split between two economic centres with each accounting for substantial shares of the AI start-ups (Johannesburg/ Cape Town and Hanoi/Ho Chi Minh City). AI Start-Ups That Closed The focus in this assessment is on continuing AI start-ups that are participating in the AI economy in our sample of middle-income developing countries. But it is worthwhile to pause and consider the AI start-ups not included in our sample because they have closed (see Table 11). Of the 133 closed AI start-ups in the 10 case study countries, Crunchbase reports that only 11 had a G7 investor, while another 26 are listed as having exclusively domestic or third country investors. For the other 96 closed AI start-ups, it appears that bootstrapping was a primary means for funding the firm (though we cannot rule out a gap in the reporting on investor participation). Of the 110 closed firms that reported employment, three quarters had employment of 10 or less. Only 30 of the 133 closed firms provided reporting on closing or exit dates. Among the closed firms, “exit” is likely the preferred way out, meaning the firm was acquired or transformed via an initial public offering (IPO). This translates into a known “success rate” among the closed firms of about 10 percent (that is, 13 firms reported an exit, though there may be additional firms in this category not reported). In addition to AI, many of the “successful” closed firms are listed as having an industry focus in one or more areas such as content creation, digital marketing, internet, business intelligence or marketing automation. Interestingly, none of the firms with G7 investors reported having an exit, meaning that their manner of closure cannot be assessed. Among all the firms reporting closure dates (excluding exits), the age tended to be relatively short (in other words, between one and three years). South Africa was an exception, due to a couple of outlier observations. Country Total Start-Ups with G7 Investment City Number of Start-Ups, by City Market Shares (%) Nigeria 22 Lagos 14 63.6 Yaba 2 9.1 Abuja 2 9.1 Other cities 4 18.2 South Africa 21 Cape Town 9 42.9 Johannesburg 8 38.1 Other cities 4 19.0 Thailand 2 Bangkok 2 100.0 Tunisia 7 Tunis 6 85.7 Sfax 1 14.3 Vietnam 12 Hanoi 6 50 Ho Chi Minh City 6 50 Total 239 Source: www.crunchbase.com/discover/organization.companies; author’s tabulations and database construction; Claude.ai (table compilation assistance). Table 10: Geographic Concentration of AI Start-Ups (Continued) 29Developing Countries’ Business Participation in the AI Economy Table 11: Assessment of AI Start-Ups that Closed Down, Counts (Except “Years” Where Age Is Reported) Home Country AI StartUps That Have Closed Firms with External Investors Listed Firms with Listed G7 Investors Employment (1–10) Employment (11–50) Employment (51–100) Employment (101–250) Employment (251–500) Employment (firms failing to report) Firms Reporting Closing Date (excluding those exiting) Average Age At Closure, Years Exit Reported: Acquired or IPO Average Age At Exit, Years Closed Firms Failing to Report Closure or Exit Date Brazil 80 26 6 49 7 3 3 1 17 92.9 11 5.9 60 Colombia 5 1 1 2 2 1 0 0 0 2 1.7 0 --- 3 Egypt 5 2 1 4 1 0 0 0 0 1 1.3 0 --- 4 Indonesia 4 1 1 2 2 0 0 0 0 2 1.2 1 1.9 1 Kenya 8 1 1 7 1 0 0 0 0 0 --- 0 --- 8 Nigeria 13 2 0 11 1 0 0 0 1 1 1.2 0 --- 12 South Africa 7 3 1 4 1 1 0 0 1 2 7.5 15.0 4 Thailand 1 0 0 1 0 0 0 0 0 0 -- 0 --- 1 Tunisia 3 0 0 3 0 0 0 0 0 0 -- 0 --- 3 Vietnam 7 1 0 1 1 0 0 1 4 0 -- 0 --- 7 Totals 133 37 11 84 16 5 3 2 23 17 13 103 Sources: Crunchbase, online database, www.crunchbase.com/discover/organization.companies (viewed at end of Q1 2025); author’s tabulations and database construction. Note: IPO = initial public offering. 30 CIGI Paper No. 323 — June 2025 • Douglas Lippoldt In Brazil (the only country with more than a couple of observations), the age of successful AI start-ups that exited tended to be older than for the average AI start-up that closed. It may be that entrepreneurs and investors there are relatively quick to pull the plug on a failing firm, even as they recognize that it takes longer to prepare an exit for a successful start-up. Unfortunately, due to gaps in the data set, we are not able to confirm this for the overall population of closed start-ups. AI Unicorns To further assess the AI start-up-investor universe in our case study countries, we now turn to look for unicorns. Such firms are interesting reference cases for other start-ups in that they have demonstrated possible pathways to success. We draw on the database of unicorns developed by CB Insights.40 This database provides information on the companies, their valuations, date of recognition as a unicorn, country and city of headquarters, applicable industry, and selected investors. As of January 7, 2025, the roster included 1,257 firms.41 The current roster of unicorns includes 37 firms based in our case study countries.42 Only one of these unicorns is also in our Crunchbase database of AI start-ups. (That is the Brazilian start-up Cloudwalk, an AI-powered financial services provider with a valuation of US$2.15 billion.) But there are other AI firms and investors of interest in the unicorn roster. In order to identify the top AI-intensive firms on the list, our technique is to tap into industry expert opinion concerning which AI firms excel and then cross reference this against the roster of unicorns.43 The industry publication eWeek releases such a list of leading AI firms, with the latest edition published on October 1, 2024 and covering 150 firms (Hiter 2024). On this basis, 16 AI-intensive firms can be identified (Table 12). Of these, 13 are located in the United States, two in the United Kingdom and one in Canada. 40 See www.cbinsights.com/research-unicorn-companies. Also note Appendix 1. 41 See www.cbinsights.com/research-unicorn-companies 42 The CB Insights roster of unicorns includes 37 firms from eight of the 10 case study countries: Brazil (18), Colombia (3), Egypt (1), Indonesia (7), Nigeria (2), South Africa (1), Thailand (3) and Vietnam (2). There are no Kenyaor Tunisia-based firms on the roster. 43 See Lippoldt (2024, 10). Another option is to cross reference the unicorn roster against relevant industry association membership lists. But such an approach may prove less than timely, as developments are evolving rapidly. This unicorn analysis is of interest for our current case study of AI start-ups for a variety of reasons. First of all, the fact that most of our case study countries are represented on the roster of unicorns indicates that it is possible under local conditions to grow a start-up to unicorn scale. Secondly, given the number of early stage AI start-ups that we have identified across the 10case study countries, it may prove relevant to keep in mind that AI start-ups can become unicorns and tend to do so earlier in their life cycle than non-AI firms. According to CB Insights,44 among unicorns emerging in 2024, nearly two thirds of AI unicorns attained the status prior to scaling up and becoming fully established. Most of these AI unicorns were still in the phases of validating their business models with the initial deployment of resources. For more than four in five non-AI unicorns, their status was attained in later phases of commercial maturity, as they scaled and established themselves. Given high market expectations of AI technology start-up businesses, there may be leverage that can be exploited in seeking investors.45 Thirdly, there is a start-up connection to the successful AI unicorns via their pool of investors. Some of these investment firms have accumulated significant experience via their collaboration with successful AI start-ups over a period of years. As noted above, a successful international engagement with such an investor can provide a channel for technology transfer, managerial skill development and operational benefits such as access to networks to tap into subject area expertise, market knowledge and suppliers. It turns out that a number of AI unicorn investors have direct ties to the AI start-ups in the case study countries. So this channel of G7 investor engagement is already operating to some extent: → Andreessen Horowitz: Invests in AI unicorns Databricks, Anduril and Shield AI. The firm also operates in Brazil with investments in two AI start-ups. → Google and Google Ventures: Invest in AI unicorns Anthropic and Synthesia. Google, Google for Startups and Google.org invest in Brazil (26 start-ups), Egypt (one start-up), 44 See www.cbinsights.com/reports/CB-Insights_AI-Report-2024.xlsx. 45 CB Insights uses a “Commercial Maturity” scoring for startups with the following phases: (1) Emerging; (2) Validating; (3) Deploying; (4) Scaling; (5) Established. 31Developing Countries’ Business Participation in the AI Economy Table 12: Leading AI Unicorns (start-ups valued at US$1bn or more), 1Q2025 Company Valuation (US$ Billions) Date Designated as Unicorn Country City Industry Select Investors OpenAI $157.00 2019-07-22 United States San Francisco Enterprise Tech Khosla Ventures, Thrive Capital, Sequoia Capital Databricks $62.00 2019-02-05 United States San Francisco Enterprise Tech Andreessen Horowitz, New Enterprise Associates, Battery Ventures Anthropic $16.05 2023-02-03 United States San Francisco Enterprise Tech Google Anduril $14.00 2019-09-11 United States Irvine Industrials Andreessen Horowitz, Founders Fund, Revolution Ventures Glean $4.60 2022-05-18 United States Palo Alto Enterprise Tech General Catalyst, Kleiner Perkins Caufield & Byers, Lightspeed Venture Partners Hugging Face $4.50 2022-05-09 United States New York Enterprise Tech Betaworks Ventures, Addition, Lux Capital Inflection AI $4.00 2022-05-13 United States Palo Alto Enterprise Tech Gates Frontier, Greylock Partners, Horizons Ventures Dataiku $3.70 2019-12-04 United States New York Enterprise Tech Alven Capital, FirstMark Capital, capitalG Shield AI $2.80 2021-08-24 United States San Diego Industrials Andreessen Horowitz, Homebrew, Point72 Ventures Moveworks $2.10 2021-06-30 United States Mountain View Enterprise Tech Lightspeed Venture Partners, Sapphire Ventures, Kleiner Perkins Caufield & Byers Synthesia $2.10 2023-06-13 United Kingdom London Enterprise Tech Google Ventures, Kleiner Perkins Caufield & Byers, FirstMark Capital Cohere $2.00 2023-05-02 Canada Toronto Enterprise Tech Index Ventures, Salesforce Ventures, Section 32 Jasper $1.50 2022-10-17 United States Austin Enterprise Tech Foundation Capital, Institutional Venture Partners, Founders Capital Runway $1.50 2023-05-04 United States New York Media & Entertainment Lux Capital, Compound, Amplify Partners Adept $1.00 2023-03-14 United States San Francisco Enterprise Tech Greylock Partners, Addition, M12 Stability AI $1.00 2022-10-05 United Kingdom London Enterprise Tech Lightspeed Venture Partners, Coatue Management Memo item CloudWalk $2.15 2021-09-08 Brazil Sao Paulo Financial Services Plug & Play Ventures, Valor Capital Group, DST Global (also Coatue, BTG Pactual) Source: CB Insights (2025a); Hiter (2024); author’s tabulations; Claude.ai for compilation. Additional investors for CloudWalk (in brackets) were identified from the Crunchbase database. Note: CloudWalk self identifies as “the interplanetary payment network.” 32 CIGI Paper No. 323 — June 2025 • Douglas Lippoldt Indonesia (one startup); Kenya (5 start-ups); Nigeria (5 start-ups); and South Africa (one startup). → Sequoia Capital: Invests in OpenAI. It is also operating in Brazil (one start-up). → Khosla Ventures: Invests in OpenAI. It is also investing in Nigeria (one start-up). → Salesforce/Salesforce Ventures: Invests in Cohere. The firm is also operating in Tunisia (one startup). → Coatue: Invests in Stability AI. It is also investing in Brazil (CloudWalk). The Brazilian start-up ecosystem appears to be fairly well integrated with ties to four of these investor groups. Nigeria has connections to two of the investors and Kenya has had success with the Google group. But connections to the others are more limited. Colombia, Thailand and Vietnam are not yet engaged via this channel, so this remains a possible opportunity for these countries to explore. Policy Assessment In light of the foregoing assessment of start-ups with G7 investment, we now turn to the policy environment in which they operate. We conducted a review of AI policy frameworks in each of the case study countries. Data from the OECD.AI Policy Observatory was employed as a baseline for the review. We supplemented this review by drawing on data from the activity tracker at the Digital Policy Alert (DPA) organization, as well as information from several DPA analytical reports.46 Other updates were drawn from peerreviewed academic and regional sources.47 The 10 case study countries all have launched development of AI policy frameworks to advance their engagement in the AI economy (see Box 5). They vary in their progress (details are given 46 For further information on the DPA as a data source, see Appendix 1. 47 The full list of sources can be found at the bottom of Table A.4 in Appendix 2. Note: A summary table view of the OECD.AI data as of 2021 can be found in Lippoldt (2024, table 5, 21–23). The table covers the advanced economies as well as the present case study countries Brazil, Indonesia, South Africa and Vietnam. in Table A.4 in Appendix 2). It is interesting to see that Colombia and Brazil — with the most developed AI policy frameworks — also have the largest numbers of AI start-ups among our case study countries (see Table 5). Both countries have taken steps toward the development of advanced regulatory mechanisms such as sandbox approaches and risk-based legislation.48 South Africa, Thailand, Tunisia and Vietnam occupy an intermediate position in AI policy development with defined national strategies, designated priorities and emerging regulatory frameworks. However, these four countries each face implementation challenges. Kenya and Nigeria show promising momentum with recent legislative developments, despite infrastructure limitations. Egypt and Indonesia represent earlier stages of policy formation, focusing primarily on capacity building and infrastructure development while still establishing basic regulatory frameworks. Across all of the case study countries, there is a common tension between fostering innovation and ensuring ethical governance. Countries are responding to this tension with engagement in international cooperation and various adaptations to the divergent regulatory approaches of their international partners. Some are leaning toward Western-aligned frameworks (in other words, the EU model) and others to alternative models influenced by partnerships with countries such as China. Policy Review Findings Drawing on the summaries above and the detailed policy data in Table A.4 in Appendix 2, several themes emerge. First, the 10 case study countries have all begun to tackle AI policy and regulation using strategic approaches. Each of the sample countries has recognized that the scope of the challenge and opportunity of AI requires a national-level response. Accordingly, they have launched initiatives to take stock and develop national strategies or plans and action agendas. Progress in these initiatives is mixed, as can be seen in Table A.4 in Appendix 2. Yet it is notable that the sample countries are each 48 With respect to AI start-ups, it may be that the consultative approaches being employed in these two nations are providing stakeholders with greater certainty and predictability in the commercial environment. This is valued by entrepreneurs and may help to promote start-up development. For example, see the discussion on certainty and predictability in regulatory processes in Geradin (2017, section C). 33Developing Countries’ Business Participation in the AI Economy striving to capitalize on AI in a strategic manner, taking into account their national conditions. Brazil and Colombia are furthest in development of their AI frameworks and have achieved some scale in development of their AI ecosystems, which they are both seeking to leverage as emerging regional players. Egypt, Kenya and Tunisia have sought to position themselves as aspiring regional AI hubs as well, albeit on a smaller scale. Several others — such as Indonesia, South Africa and Thailand — refer to development of AI as part of strategic engagement in the socalled Fourth Industrial Revolution, which refers to positive exploitation of technologies that can leverage inputs from the physical, digital and biological spheres (for example, with respect to automation or cyber-physical systems). Second, most of the countries are facing implementation challenges in realizing their national plans. This is, in part, associated with developing adequate capacity in public administration and among other stakeholders Box 5. Case Study Countries’ AI Policy Framework Status in a Nutshell, 1Q2025 Brazil: Fairly comprehensive framework for robust regulatory development. The system is oriented towards EU-type approaches to AI governance including a recently passed Senate AI bill focused on risk management (lower house action is pending) and a proposed National Center for Algorithmic Transparency and Trustworthy AI. Colombia: Extensive policy initiatives, including some with regulatory sandboxes through the Superintendencia de Industria y Comercio’s innovation lab; balancing innovation with responsible use through a comprehensive pillar-based strategy. Egypt: Strategy focused on becoming a regional AI hub through capacity building and research, with growing Chinese partnerships potentially creating Western alignment challenges. Indonesia: Early-stage framework prioritizing infrastructure development across five priority areas, with recently implemented data protection legislation and developing ethics guidelines. Kenya: Emerging strategy emphasizing key economic sectors with strong mobile infrastructure and active international governance engagement but facing resource constraints. Regulatory system is still in early stages. Nigeria: Evolving framework with emphasis on ethics and inclusivity; significant legislative progress with new regulatory commission approval, despite persistent infrastructure challenges. South Africa: Balanced approach focusing on inclusive growth and talent development, with the most developed venture capital ecosystem in Africa, but hampered by chronic electricity supply challenges. Thailand: Well-developed risk-based strategy with detailed targets and implementation plan; actively participating in ASEAN governance frameworks while developing sector-specific applications. Tunisia: Strategy focused on leveraging engineering education to become a regional AI hub, with an EU-aligned data protection framework, but facing a challenge from implementation gaps and limited enforcement. Vietnam: Forward-looking approach with regulatory innovation including machine-readable AI labels and sandboxes; developing substantial specialist workforce amid growing partnerships with Asian technology leaders. Sources: See Table A.4. 34 CIGI Paper No. 323 — June 2025 • Douglas Lippoldt that need to be engaged. Some have sought to meet this challenge via public-private partnership arrangements.49 In developing the regulatory regimes associated with implementation of the national plans, some of the case study countries have also underscored the need for transparency and consultation, which can have positive effects on outcomes. For example, Brazil, Colombia, Kenya, South Africa and Thailand all have made explicit references to these issues with respect to specific aspects of regulation. Third, all 10 of the countries are still struggling with the urban-rural digital divide, a key element for inclusiveness in the AI economy. Closing the gap will require — among other elements — improved performance in education and infrastructure development (for example, electricity, telephony, information technology capacity). Fourth, nearly all of the case study countries have established targets for AI human capital development.50 Shortfalls are proving to be a constraint on AI development, not least because of international poaching of talent. Brazil, Egypt and Tunisia explicitly cite talent flight as a strategic risk. (And, indeed, Maslej et al. (2024, 239) report that countries such as Brazil and South Africa have experienced a steady outflow of talent to competitor markets.) Fifth, there is a tension over regulatory convergence and divergence, with implications for market access in both directions for the case study countries. Selling and sourcing are both affected by regulatory compliance issues. There are risks due to fragmentation. Brazil, Colombia, South Africa and Tunisia have taken steps toward EU style approaches to regulation, while Egypt, Kenya, and Vietnam have begun collaboration with some Chinese AI initiatives (for example, via the Belt and Road Initiative (BRI), which includes a digital dimension). The United Kingdom as well has an influence, in particular on AI safety issues. Moreover, the US market is a key source of AI technology and venture capital for all 10 countries as demonstrated in our start-up analysis. Early 49 As noted in the literature review, Mutasa et al. (2024) point to publicprivate partnerships as one means for stakeholders to mitigate high-cost burdens (for example, with respect to infrastructure projects). 50 It is notable that action in this area is also identified by Fan and Qiang (2024), who emphasize prioritizing skills development as one of five priorities in the World Bank framework for positive AI development (see the literature review above for a discussion of this reference). indications are that the US is now shifting to a much lighter regulatory regime (Carrillo et al. 2025; Villasenor and Turner 2024). While the terms of the new US AI regime still remain unsettled, there are risks of US policy changes further fracturing global AI governance. For the 10 case study countries, alignment with any one of the AI major economies could have knock-on effects in terms of market access and compliance with the other major economies. This is particularly important with respect to access to and control of data, key ingredients for the AI economy. It could affect other areas as well, such as via compliance issues with respect to exports of equipment or cross-border sales of AI services. Conclusions This case study considers the situation of a sample of internationally engaged AI start-ups and the role of the state in shaping the economic context for firm-level development. The research has identified some 2,537 AI start-ups in the 10 case study countries operating across a broad range of sectors. Drawing on a combination of global and domestic inputs and innovation, these firms are identifying local and regional needs and developing AI-supported solutions for commercial markets. In doing so, about one in 10 of these firms have succeeded in attracting investment from entities based in the G7 country group. Such investment provides a channel for transfer of knowhow concerning development and commercial deployment of the technology beyond what might be available domestically. Domestic policy is playing a complementary role to this commercial activity. From build out of infrastructure to human capital development, and regulation, national governments are establishing the context and conditions in which the AI ecosystem is operating. As in much of the world, this is a work in progress in each of the ten case study countries. Small Scale, Bigger Importance The scale of the AI startup activity in the study countries is small compared to the scale of comparable activity in the leading AI nations such as the United States, China or Europe and the United Kingdom. But it is notable in that entrepreneurs in the study countries have 41Developing Countries’ Business Participation in the AI Economy Table A.2: The Top AI-Start-Up Investor Based in Each Case Study Country, Counts of Investment Engagements by Destination Country, as of Q1 2025 Investor Name Headquarters Brazil Colombia Egypt Indonesia Kenya Nigeria South Africa Thailand Tunisia Vietnam Total Investment Style Support Offerings Bossa Invest Brazil 15 0 0 0 1 0 1 0 0 0 17 Early-stage seed investor with AI specialization; portfolio spans multiple sectors with domestic focus while opportunistically exploring African markets. Provides hands-on management guidance and strategic direction; offers operational support to portfolio companies with limited technical assistance. Latin Leap Colombia 0 1 0 0 0 0 0 0 0 0 1 Venture Capital Studio focused on soft-landing purposedriven tech companies in Latin America; emphasizes ecosystem development over pure returns. Offers network access and regional connections; positions itself as a gateway for international companies entering the Colombian market. Flat6Labs Egypt 0 0 2 0 0 0 0 0 0 0 2 Leading MENA region seed accelerator with structured program; focuses on supporting early-stage Egyptian start-ups with regional growth potential. Capital-focused investor with program-based support; offers standardized accelerator curriculum rather than customized assistance. MDI Ventures Indonesia 0 0 0 3 0 0 0 0 0 0 3 Multi-stage venture fund with unicorn track record; corporate venture arm of Telkom Indonesia with both strategic and financial objectives. Primarily financial investor with limited operational support; leverages parent company's regional business connections for portfolio companies. Catalyst Fund Kenya 0 0 1 0 1 0 0 0 0 0 2 Impact-focused fintech accelerator targeting underserved populations; combines financial inclusion mission with market-based approach. Provides network access to investors and partners; emphasizes ecosystem connections over direct technical or management support. iNOVO Nigeria 1 2 1 0 0 0 0 0 0 0 4 Accelerator program powered by UK-Nigeria Tech Hub partnership; focuses on earlystage start-ups across multiple sectors with regional outlook. Offers primarily capital and program structure; limited ongoing support beyond initial acceleration phase. Injini South Africa 0 0 0 0 3 0 0 0 0 0 3 Africa's first EdTech incubator and seed investor; specialized focus on educational technology with pan-African investment strategy. Capital focused with programbased support; offers standardized incubation rather than customized technical or management assistance. Flat6Labs Tunisia 0 0 4 0 0 0 0 0 1 0 5 Regional seed-stage VC firm operating across MENA; structured accelerator programs combined with seed funding for early-stage ventures. Provides technical expertise and resources; offers standardized startup support program with emphasis on technical development. 500 Startups Vietnam Vietnam 0 0 0 0 0 0 0 0 0 2 2 Tech-focused seed investor affiliated with global 500 Startups network; combines local expertise with international investment approach. Offers technical expertise and global connections; provides standardized accelerator support with technical resources for early-stage companies. Source: www.crunchbase.com/discover/organization.companies; author’s tabulations and investor database construction; Claude.ai (final table compilation assistance). Note: The firm descriptions draw heavily on each firm’s self-characterization and public information. Thailand had no local top five investors in sample AI start-ups. 42 CIGI Paper No. 323 — June 2025 • Douglas Lippoldt Table A.3: The Overall Top 10 AI-Start-Up Investors Targeting the Case Study Countries, Counts of Investment Engagements by Destination Country, as of Q1 2025 Investor Name Headquarters Brazil Colombia Egypt Indonesia Kenya South Africa Thailand Tunisia Vietnam Nigeria Total Investment Style Support Offerings Google for Startups United States 26 0 0 1 4 1 0 0 0 4 36 Early-stage accelerator program that creates global connections for start-ups; specializes in software and digital platforms. Offers mentorship, tech resources, Google product access and ecosystem networking opportunities but limited direct management support. Bossa Invest Brazil 15 0 0 0 1 1 0 0 0 0 17 Early-stage seed investor with strong domestic focus; industry-agnostic but with AI specialization. Provides hands-on management guidance and active involvement in strategic decisions; limited technical assistance. Canary Brazil 12 0 0 0 0 0 0 0 0 0 12 First-check Latin American seed fund targeting innovative tech companies with scalable business models; strong domestic focus. Capital-focused investor with limited operational involvement; offers network connections within Brazilian tech ecosystem. Norte Ventures Brazil 10 0 0 0 0 0 0 0 0 0 10 Pure-follower fund created and managed by entrepreneurs; focuses on early-stage Brazilian start-ups. Primarily financial investor without significant value-added support; relies on lead investors for operational guidance. Y Combinator United States 5 0 1 3 0 0 0 0 0 1 10 A leading global start-up accelerator with standardized investment terms; batchoriented program with strong alumni network. Provides structured startup curriculum, demo day exposure and founder network access; limited ongoing management support. Techstars United States 2 0 0 0 2 1 0 0 1 3 9 Global accelerator with regional programs organized by industry verticals; highly structured three-month program. Comprehensive support including mentorship network, technical resources and business development assistance for early-stage growth. Latitud Ventures Brazil 8 0 0 0 0 0 0 0 0 0 8 Community-focused investment platform aimed at Latin American founders; combines capital with infrastructure. Offers founder community, technical infrastructure and networking resources for early-stage companies. Amazon Web Services United States 5 0 1 0 0 0 0 0 1 0 7 Corporate investor prioritizing cloud adoption and technical ecosystem development; strategic investment approach. Provides substantial technical support through cloud credits, architecture assistance and technology integration expertise. Village Capital United States 2 0 0 0 2 1 0 0 0 2 7 Impact-focused investor using peer-selection model; targets underserved markets and founders. Combines capital with business curriculum, peer mentoring and investor connections for social impact ventures. 500 Global United States 1 0 1 1 0 0 1 0 2 0 6 Early-stage global VC with regional funds; high-volume investment strategy across diverse geographies. Limited operational support but offers global network access and follow-on funding potential; regional offices provide local expertise. Startupbootcamp United Kingdom 1 2 2 0 0 1 0 0 0 0 6 Industry-focused global accelerator network with structured programs; operates regionally with concentrated cohorts in specific markets. Comprehensive accelerator model with formal mentorship program, networking opportunities and industry connections for portfolio companies. Source: www.crunchbase.com/discover/organization.companies; author’s tabulations and investor database construction; Claude.ai (final table compilation assistance). Notes: The firm descriptions draw heavily on each firm’s self-characterization and public information. The top 10 listing includes 11 companies due to a tie in the counts.. 43Developing Countries’ Business Participation in the AI Economy Table A.3: The Overall Top 10 AI-Start-Up Investors Targeting the Case Study Countries, Counts of Investment Engagements by Destination Country, as of Q1 2025 Investor Name Headquarters Brazil Colombia Egypt Indonesia Kenya South Africa Thailand Tunisia Vietnam Nigeria Total Investment Style Support Offerings Google for Startups United States 26 0 0 1 4 1 0 0 0 4 36 Early-stage accelerator program that creates global connections for start-ups; specializes in software and digital platforms. Offers mentorship, tech resources, Google product access and ecosystem networking opportunities but limited direct management support. Bossa Invest Brazil 15 0 0 0 1 1 0 0 0 0 17 Early-stage seed investor with strong domestic focus; industry-agnostic but with AI specialization. Provides hands-on management guidance and active involvement in strategic decisions; limited technical assistance. Canary Brazil 12 0 0 0 0 0 0 0 0 0 12 First-check Latin American seed fund targeting innovative tech companies with scalable business models; strong domestic focus. Capital-focused investor with limited operational involvement; offers network connections within Brazilian tech ecosystem. Norte Ventures Brazil 10 0 0 0 0 0 0 0 0 0 10 Pure-follower fund created and managed by entrepreneurs; focuses on early-stage Brazilian start-ups. Primarily financial investor without significant value-added support; relies on lead investors for operational guidance. Y Combinator United States 5 0 1 3 0 0 0 0 0 1 10 A leading global start-up accelerator with standardized investment terms; batchoriented program with strong alumni network. Provides structured startup curriculum, demo day exposure and founder network access; limited ongoing management support. Techstars United States 2 0 0 0 2 1 0 0 1 3 9 Global accelerator with regional programs organized by industry verticals; highly structured three-month program. Comprehensive support including mentorship network, technical resources and business development assistance for early-stage growth. Latitud Ventures Brazil 8 0 0 0 0 0 0 0 0 0 8 Community-focused investment platform aimed at Latin American founders; combines capital with infrastructure. Offers founder community, technical infrastructure and networking resources for early-stage companies. Amazon Web Services United States 5 0 1 0 0 0 0 0 1 0 7 Corporate investor prioritizing cloud adoption and technical ecosystem development; strategic investment approach. Provides substantial technical support through cloud credits, architecture assistance and technology integration expertise. Village Capital United States 2 0 0 0 2 1 0 0 0 2 7 Impact-focused investor using peer-selection model; targets underserved markets and founders. Combines capital with business curriculum, peer mentoring and investor connections for social impact ventures. 500 Global United States 1 0 1 1 0 0 1 0 2 0 6 Early-stage global VC with regional funds; high-volume investment strategy across diverse geographies. Limited operational support but offers global network access and follow-on funding potential; regional offices provide local expertise. Startupbootcamp United Kingdom 1 2 2 0 0 1 0 0 0 0 6 Industry-focused global accelerator network with structured programs; operates regionally with concentrated cohorts in specific markets. Comprehensive accelerator model with formal mentorship program, networking opportunities and industry connections for portfolio companies. Source: www.crunchbase.com/discover/organization.companies; author’s tabulations and investor database construction; Claude.ai (final table compilation assistance). Notes: The firm descriptions draw heavily on each firm’s self-characterization and public information. The top 10 listing includes 11 companies due to a tie in the counts.. Table A.4: AI Policy Overview in Case Study Countries Brazil Colombia National AI Strategies and Agendas Its national AI strategy (2021) has pillars for research, governance, workforce development and international cooperation. Current AI plan 2024–2028 proposes a National Centre for Algorithmic Transparency and Trustworthy AI. Comprehensive AI bill passed Senate in December 2024; lower house approval pending. Established National Policy for Digital Transformation and AI in 2019 with significant government commitment to implementation. The policy is comprehensive with strategic pillars addressing technology adoption barriers, innovation conditions, human capital development and AI preparedness. Access to AI Development Finance Emerging venture capital ecosystem is concentrated in São Paulo, with international investors increasingly active. Some government grants available. International funding access may be impacted by diverging regulatory approaches between Brazil and major markets such as the United States, potentially complicating cross-border flows. Government funding is available via its entrepreneurship agency including a five-year plan and budget allocation for AI; partnerships with international organizations are boosting this. CEmprende initiative provides support for AI start-ups; innovation centres in cities are also fostering a start-up ecosystem. Emerging Regulatory Framework Privacy legislation implemented in 2020 provides data protection foundation. Brazil’s proposed AI regulation adopts a risk-based approach similar to the EU AI Act, requiring algorithmic impact assessments and human oversight for high-risk systems. The bill uniquely mandates compensation for copyright holders when their content is used to train AI systems. Risk of US regulatory clash. Regulatory sandbox approach through the Superintendency of Industry and Commerce’s (Colombia’s national regulatory agency) innovation lab. Data protection regime is being established with emerging AI-specific ethical guidelines for AI deployment. Seeking to balance an enabling environment for innovation against need for responsible AI use. AI Education, Training and Skills Development An AI talent pipeline is developing at leading universities. National digital skills program aims to train 100,000 professionals in digital technologies. But there is a risk of talent flight abroad. “Misión TIC 2022” program aims to train 100,000 programmers. University programs in AI developing with international academic partnerships. International AI Cooperation Active participation in OECD AI initiatives and bilateral agreements with European Union on AI development. Member of Global Partnership on AI. Signed the Bletchley Declaration on AI Safety. Increasing regulatory alignment with the EU model. Active engagement with OECD AI Principles and partnerships with countries such as Canada and South Korea on AI development. Such cooperation is pursued to ensure that Colombia’s AI stance is competitive. Infrastructure Readiness Uneven digital infrastructure with high connectivity in urban centres, some lag in rural areas. Cloud computing infrastructure developing rapidly. Data localization requirements mandate cloud service providers store local copies of government data in Brazil, potentially posing a challenge for international AI services. Significant investment in digital infrastructure is under way with the national broadband plan, though rural connectivity remains challenging. Bridging the urban-rural digital divide is a priority area. Implementation Challenges Compliance challenges from uncertainty on enforcement mechanisms and the future regulatory authority’s structure. Implementation is affected by political transitions. Public sector inefficiencies and AIhesitancy from SMEs are constraints. Industry-Specific AI Applications Strong focus on agricultural AI applications, financial services and natural resources in Brazil’s economic priorities. Emerging focus on consumer protection in AI systems, with active enforcement. Policy focus is on public services, health-care diagnostics and agricultural applications tailored to development priorities. Also, there is an AI role to improve rural education access and financial inclusion. International Regulatory Divergence Trump administration’s new approach to AI regulation, with deregulation and heightened export controls, may pose risk to market access for Brazilian startups. Also, Brazil’s alignment with EU-style regulation may increase operational costs for start-ups. Colombia is working toward harmonization with international frameworks, particularly OECD principles, which could help mitigate regulatory divergence. US policy uncertainty poses a risk. 44 CIGI Paper No. 323 — June 2025 • Douglas Lippoldt Table A.4: Case Study Country, AI Policy Overview (Continued) Egypt Indonesia National AI Strategies and Agendas National AI strategy launched in 2021 focusing on capacity building, research, ethics and governance with ambition to become regional AI hub. National AI strategy (Stranas KA) released in 2020 with five priority areas including health, bureaucratic reform, education, food security and mobility. Access to AI Development Finance The Information Technology Industry Development Agency provides grant support for innovation. International investors showing interest in Egyptian tech start-ups, though AI-specific funding remains limited. VC activity growing in Jakarta. Government’s 1,000 Start-up Movement is providing some support for early-stage companies including AI start-ups. Emerging Regulatory Framework Data Protection Law of 2020 provides foundation. National AI ethics charter under development to guide responsible AI deployment. Personal data protection legislation recently implemented. AI ethics guidelines under development through Ministry of Research and Technology. AI Education, Training and Skills Development Specialized AI faculties established at universities and Egyptian AI centre for education. Partnership with international tech companies for workforce training. Digital talent gap addressed through National Digital Talent Scholarship program. University partnerships with global tech companies developing AI curriculum. International AI Cooperation Regional partnerships within Arab League on AI governance. Growing collaboration with Chinese AI initiatives through BRI ASEAN Digital Innovation Network participation. Bilateral AI cooperation with Singapore and Japan. Infrastructure Readiness Uneven digital infrastructure with significant investments in smart city technologies but challenges in broader connectivity. Archipelagic geography creating digital infrastructure challenges. Palapa Ring project improving connectivity across islands. Implementation Challenges Bureaucratic processes slowing implementation. Brain drain of technical talent to Gulf states and Western countries. Policy implementation varies significantly across different regions. Coordination between multiple government agencies remains challenging. Industry-Specific AI Applications Focus on Arabic natural language processing, health-care diagnostics and archaeological applications reflecting Egypt’s specific context. Marine resource management, disaster prediction systems and financial inclusion reflect Indonesia’s specific geographic and development context. International Regulatory Divergence Egypt’s approach to AI regulation shows potential divergence with Western frameworks through its growing collaboration with Chinese AI initiatives via BRI partnerships. This may create regulatory alignment challenges with Western markets as Egypt positions itself as a regional AI hub while balancing international partnerships. Indonesia is working to align its regulatory approach with ASEAN frameworks while also developing partnerships with Singapore and Japan. However, its personal data protection framework may face challenges harmonizing with stricter regimes such as the EU General Data Protection Regulation, potentially affecting international data flows and AI applications that rely on cross-border data sharing. 45Developing Countries’ Business Participation in the AI Economy Table A.4: Case Study Country, AI Policy Overview (Continued) Kenya Nigeria National AI Strategies and Agendas National AI strategy emphasizes agriculture, health care, manufacturing and housing with a focus on leveraging Kenya’s position as a regional tech hub. Consultation on the strategy finished in January 2025. The draft promotes implementation in manufacturing, financial sectors and other priority areas. “Silicon Savannah” tech hub aims to foster an innovation ecosystem. National Digital Economy Policy and Strategy (2020–2030) has AI components. National Center for AI and Robotics established to coordinate development. In August 2024, the draft national AI strategy was published outlining guiding principles including ethics, inclusivity, transparency and risk management. Digital economy: 11.30 percent of GDP, Q3 2024, up 14 percent year-on-year. Access to AI Development Finance VC interest in Kenyan tech start-ups is growing. International donors provide AI-for-development funding. World Bank approved US$390 million in 2023 for acceleration project. VC interest in Nigerian tech start-ups is ongoing, particularly financial technology (fintech). Some government funding through National Digital Innovation and Entrepreneurship Fund. Emerging Regulatory Framework The data protection law (2019) provides a foundation. The Task Force on Blockchain and AI is developing ethical guidelines. Kenya Bureau of Standards published the Draft Information Technology AI Code of Practice in April 2024 to guide responsible AI development. Kenya Robotics and AI Society Bill introduced in 2023 aims to create regulatory body for AI sector. Kenya Open Data Initiative facilitates AI development using government data. Nigeria’s Data Protection Regulation (2019) provides a foundation. National AI ethics guidelines are under development by the National Center for AI and Robotics (NCAIR). In December 2024, the House of Representatives passed a bill to establish the National Institute for Artificial Intelligence and Robotics Studies Regulation Commission, incorporating several proposals including the Control of Usage of AI Technology Bill. AI Education, Training and Skills Development AI academic programs are linking Strathmore and University of Nairobi. IBM Research Africa and Google AI Lab in Nairobi are building research capacity. Bilateral agreements with United States and other nations are focusing on digital upskilling and AI capacity building. AI education initiatives at leading universities. Private sector training programs by companies such as Data Science Nigeria addressing formal education gaps. NCAIR developing capacity building programs. International AI Cooperation Participation in UN AI for Good initiatives. Partnership with the United Kingdom on responsible AI development through Digital Access Programme. Kenya joined Paris Charter on AI in February 2025 and participates in various international AI governance initiatives including the Seoul Declaration on AI Safety and the Bletchley Declaration. Limited formal international AI partnerships, though engaged with pan-African AI initiatives. Nigeria has signed international AI governance frameworks including the Bletchley Declaration on AI Safety (2023), the Paris Charter on AI (2025) and the African Union’s Continental AI Strategy and AI Governance Framework. Infrastructure Readiness Kenya enjoys strong mobile connectivity infrastructure. High mobile money penetration provides a foundation for AI fintech. By Q4 2024, mobile network connections reached US$66.1 million with 128.3 percent penetration rate, enabling a robust digital ecosystem. Electricity infrastructure challenges affect digital deployment. Mobile connectivity growing; broadband access remains limited. Information and communications sector increasingly important: 142 million active internet subscribers by January 2025. Implementation Challenges Resource constraints for policy implementation. Urban-rural digital divide persists. Regulatory system for AI still in early stages. Implementation affected by resource constraints and coordination challenges (federal/state). Early stage of AI regulation rollout. Industry-Specific AI Applications Mobile-based agricultural advisory services, financial inclusion technologies and wildlife conservation reflect economic priorities. AI implementation promoted in manufacturing and finance. Financial inclusion technologies, health -are diagnostics and natural language processing for Nigeria’s diverse linguistic landscape. Potential for growth in fintech, strong VC interest. International Regulatory Divergence Kenya balancing African Union Continental AI Strategy, cooperation with China on AI research, EU-Smart Africa cooperation and Westernled initiatives such as the Paris Charter. Nigeria is navigating multiple international AI frameworks with different approaches, as well as the African Union and the African Continental Free Trade Area’s Digital Trade Protocol. 46 CIGI Paper No. 323 — June 2025 • Douglas Lippoldt Table A.4: Case Study Country, AI Policy Overview (Continued) South Africa Thailand National AI Strategies and Agendas Presidential Commission on 4IR recommendations for AI adoption. National AI Policy Framework published in August 2024 focusing on inclusive economic growth, talent development, infrastructure, innovation and ethical AI. AI Strategy (2022–2027) focuses on competitiveness and workforce development. Thailand 4.0 policy positions AI within broader digital transformation. The national AI strategy includes an action plan to advance AI development across sectors including agriculture, health care and finance, with a focus on building human resources and fostering innovation. Access to AI Development Finance Most developed venture capital ecosystem in Africa. Technology Innovation Agency providing government support for deep tech start-ups. Thailand has a growing start-up ecosystem with increasing venture capital activity. Government support is available through Digital Economy Promotion Agency and innovation funds. Emerging Regulatory Framework Protection of Personal Information Act provides data protection foundation. AI ethics frameworks under development through Department of Science and Innovation. National Data and Cloud Policy adopted in March 2024 with provisions affecting AI development. The Personal Data Protection Act provides a data protection foundation. AI ethical guidelines being developed through Digital Economy and Society Ministry. A draft Royal Decree on Business Operations Using AI Systems adopts a risk-based approach, categorizing AI systems into prohibited and high risk. AI Education, Training and Skills Development Academic AI programs at leading universities. The Council for Scientific and Industrial Research provides additional research capacity. National Skills Fund supports AI workforce development. AI education programs are in place at leading universities. Digital workforce development is ongoing through Thailand Massive Open Online Course platform International AI Cooperation Active participation in BRICS cooperation on AI governance. Engagement with OECD AI principles despite non-member status and African Union’s Continental AI Strategy endorsed in June 2024 Thailand participates in ASEAN’s AI governance framework and has bilateral AI development partnerships with Japan and Singapore. Infrastructure Readiness Most advanced digital infrastructure in Sub-Saharan Africa, though significant urban-rural divide remains. Investment in telecommunication networks (approx. US$10.6bn) and data centres (approx. US$1.1bn) over the past five years. Chronic electricity supply challenges, including load shedding and power outages, create significant barriers for AI infrastructure reliability, increasing operational costs for AI start-ups and potentially limiting compute-intensive AI applications. There is a strong digital infrastructure in urban centres, with the Thailand Digital Valley project expanding capacity. A national broadband network is under development. Thailand has also been working to improve electrical electricity grid reliability, which varies significantly between urban and rural areas; it is deploying smart grid technology and energy storage solutions. Implementation Challenges Economic inequality affecting equitable AI deployment. Policy implementation delayed by consultation processes. Coordination challenges between multiple agencies involved in digital policy. Skills gap affecting implementation of advanced AI. Industry-Specific AI Applications Per South Africa's economic structure and social priorities: mining safety and efficiency, financial services and health-care diagnostics:. Thailand’s economic priorities for AI: medical tourism diagnostics, agricultural monitoring systems, tourism management International Regulatory Divergence Positioned between EU-style regulatory approaches and innovation-focused models. Group of 20 presidency for 2025 creates opportunity to shape global AI governance while balancing domestic development needs with international standards alignment. Thailand’s approach seeks to balance innovation with regulatory oversight through emerging frameworks such as AI sandboxes and testing centres; there are potential compliance challenges for global AI companies operating across different regulatory regimes. 47Developing Countries’ Business Participation in the AI Economy Table A.4: Case Study Country, AI Policy Overview (Continued) Tunisia Vietnam National AI Strategies and Agendas National AI strategy (2018) positions Tunisia as a regional AI hub with focus on building upon a strong engineering education system. Memorandum signed in 2022 between four ministries to outline the development and implementation of the national AI strategy. National Strategy for AI Research, Development and Application through 2030 launched in 2021 with aim to become regional leader in AI. The government advancing its Digital Economy through the National Digital Transformation Program. Access to AI Development Finance Innovation support available through Tunisian Startup Act framework. International donors providing additional support for innovation ecosystem. Growing venture capital interest in Vietnamese tech startups. Government-innovation funding through National Technology Innovation Fund and targeted AI programs. Emerging Regulatory Framework Data protection law in place since 2004. Ministry of Technology developing AI-specific regulations with focus on ethics and human rights. National Authority for the Protection of Personal Data established, though with limited regulatory authority. Cybersecurity law and data protection regulations provide a foundation. Ministry of Science and Technology developing AI governance framework. Ministry of Information and Communications published draft Law on Digital Technology Industry to regulate AI systems based on risk levels (July 2024). AI Education, Training and Skills Development A focus on developing a strong engineering education providing the foundation for AI skills. National AI capacity building program targeting 1,000 AI specialists for 2025. AI education programs at leading universities. National emphasis on STEM education providing talent pipeline with government target of 50,000 AI specialists by 2030. International AI Cooperation Mediterranean AI cooperation through partnerships with European countries. Engagement with francophone AI research networks. Tunisia signed Council of Europe’s Convention 108 for data protection in 2017. Growing partnerships with Singapore and South Korea on AI development. Increasing engagement with Chinese AI initiatives. Infrastructure Readiness Relatively advanced digital infrastructure for the region, though investment needed for advanced computing capabilities. Improving digital infrastructure and connectivity; high smartphone penetration. National program for digital transformation. Implementation Challenges Political transitions are affecting policy continuity. Economic challenges limiting domestic investment capacity. Regulatory environment still developing to keep pace with technological adoption. Talent retention challenging with global demand for Vietnamese engineers. Industry-Specific AI Applications Textile industry optimization, olive oil production monitoring and archaeological research reflect Tunisia’s economic and cultural context. AI applications enhancing Public Finance Management Information System to detect fraud and improve budget efficiencies. Based on Vietnam’s economic priorities and urbanization challenges: manufacturing optimization, aquaculture monitoring systems and urban traffic management reflecting. International Regulatory Divergence Tunisia’s data protection framework shows alignment with EU standards through adoption of GDPR principles in draft legislation and signing of Convention 108, but implementation gaps and limited enforcement authority create regulatory uncertainty for international AI developers and investors. Vietnam’s emerging risk-based AI governance has unique features prohibiting systems that manipulate behaviour without user awareness or classify individuals based on sensitive inferences. The draft Law on Digital Technology Industry requires machine-readable labels for AI-generated content, employs sandboxes. Sources: Basic information was extracted from https://oecd.ai/en/dashboards/overview, supplemented by data from Digital Policy Alert Activity Tracker online policy database; and https://globalailaw.com/colombia-ai-policy/; Buza, Jenzer and Bossard (2024); Buza and Scheiwiler (2024); Buza and Taha (2025a, 2025b); Deeg and Pierotic (2024); Giardini, Scheiwiler and Buza (2024); Lippoldt (2024); Filgueiras and Junquilho (2023); Martins (2025); Villasenor and Turner (2024); Carrillo et al. (2025); OECD (2024, chapter 2); Wadipalapa et al. (2024); República de Colombia (2024); https:// menaobservatory.ai/en/regional/20; Élysée (2025). The author’s compilation of this table was supported by Claude.ai. Note: The Paris Charter is an accord among 10 nations establishing principles for openness, accountability and participation in AI governance (Élysée 2025). 48 CIGI Paper No. 323 — June 2025 • Douglas Lippoldt Appendix 3: OECD Recommendation of the Council on Artificial Intelligence The OECD principles on AI within the OECD’s Recommendation of the Council on Artificial Intelligence1 state that: → AI should benefit people and the planet by driving inclusive growth, sustainable development and well-being. → AI systems should be designed in a way that respects the rule of law, human rights, democratic values and diversity, and they should include appropriate safeguards — for example, enabling human intervention where necessary— to ensure a fair and just society. → There should be transparency and responsible disclosure around AI systems to ensure that people understand when they are engaging with them and can challenge outcomes. → AI systems must function in a robust, secure and safe way throughout their lifetimes, and potential risks should be continually assessed and managed. → Organizations and individuals developing, deploying or operating AI systems should be held accountable for their proper functioning in line with the above principles. 1 See OECD (2019). Also note that as of March 25, 2025, the OECD legal instruments online database reports that adherence to the recommendation now covers all 38 OECD member countries and the European Union, plus nine others: Argentina, Brazil, Egypt, Malta, Peru, Romania, Singapore, Ukraine and Uruguay. The OECD recommends that governments: → facilitate public and private investment in R&D to spur innovation in trustworthy AI; → foster accessible AI ecosystems with digital infrastructure and technologies, and mechanisms to share data and knowledge; → create a policy environment that will open the way to the deployment of trustworthy AI systems; → equip people with the skills for AI, as well as support workers to ensure a fair transition; and → cooperate across borders and sectors to share information, develop standards and work toward the responsible stewardship of AI 49Developing Countries’ Business Participation in the AI Economy Works Cited Aaronson, Susan Ariel. 2024a. The Age of AI Nationalism and Its Effects. CIGI Paper No. 306. Waterloo, ON: CIGI. www.cigionline.org/publications/ the-age-of-ai-nationalism-and-its-effects/ . ———. 2024b. “Data Governance Is Not Ready for AI.” In The Role of Governance in Unleashing the Value of Data, 15–23. Waterloo, ON: CIGI. www.cigionline. org/articles/data-governance-is-not-ready-for-ai/. Abungu, Cecile, Marie Victoire Iradukunda, Duncan Cass Beggs, Aquila Hassan and Raqda Sayidali. 2024. Why Global South Countries Need to Care About Highly Capable AI. CIGI Paper No. 311. Waterloo, ON: CIGI. www.cigionline.org/publications/why-global-southcountries-need-to-care-about-highly-capable-ai/. Alfen, Nora. 2024. “Understanding the Stages of Startup Funding.” Wayra (blog), May 15. www.wayra.de/ blog/understanding-the-stages-of-startup-funding. Amodei, Dario. 2024. “Machines of Loving Grace. Section 3. Economic Development and Poverty.” October. www. darioamodei.com/essay/machines-of-loving-grace. Asia-Pacific Economic Corporation. 2024. “Digital Economy Partnership Agreement.” Submission by Chile. 2024/SOM3/CTI/004. https://mddb.apec.org/ Documents/2024/CTI/CTI3/24_cti3_004_r.pdf. Atanasovska, Daniela and Lejla Robeli. 2025. “Brazil’s AI Act: A New Era of AI Regulation.” GDPR Local, February 26. https:// gdprlocal.com/brazils-ai-act-a-new-era-of-ai-regulation/. Beckley, Michael. 2025. “The Age of American Unilateralism: How a Rogue Superpower Will Remake the Global Order.” Foreign Affairs, April 16. www.foreignaffairs. com/united-states/age-american-unilateralism. Bommasani, Rishi and Percy Liang. “Reflections on Foundation Models.” Stanford University Human-Centered Artificial Intelligence. October 18. https://hai.stanford. edu/news/reflections-foundation-models. Bremmer, Ian and Mustafa Suleyman. 2023. “The AI Power Paradox: Can States Learn to Govern Artificial Intelligence — Before It’s Too Late?” Foreign Affairs 102 (5): 26–43. www.foreignaffairs. com/world/artificial-intelligence-power-paradox. Buza, Maria, Philine Jenzer and Svenja Bossard. 2024. “Digital Policy Alert Digital Digest: Thailand.” October. https:// digitalpolicyalert.org/digest/dpa-digital-digest-thailand. Buza, Maria and Luca Scheiwiler. 2024. “Digital Policy Alert Digital Digest: Vietnam.” October. https:// digitalpolicyalert.org/digest/dpa-digital-digest-vietnam. Buza, Maria and Sherif Taha. 2025a. “Digital Policy Alert Digital Digest: Kenya.” March. https://digitalpolicyalert. org/digest/dpa-digital-digest-kenya. ———. 2025b. “Digital Policy Alert Digital Digest: Nigeria.” March. https://digitalpolicyalert.org/ digest/dpa-digital-digest-nigeria. Carrillo, Pablo E., Martin J. Mackowski, Wolfgang A. Maschek, Beth L. Goldstein, Julia B. Jacobson, Alan L. Friel and Matthew Kirk. 2025. Key Insights on President Trump’s New AI Executive Order and Policy & Regulatory Implications. Squire Patton Boggs. February. www.squirepattonboggs.com/ en/insights/publications/2025/02/keyinsights-on-president-trumps-new-ai-executiveorder-and-policy-regulatory-implications. CB Insights 2021. The Top 12 Reasons Startups Fail. www.cbinsights. com/reports/CB-Insights_Top-Reasons-Startups-Fail.pdf. Crawford, Roderick and Mark Smith. 2022. How To Solve the Data Science Skills Shortage. SAS Institute. www.sas.com/ content/dam/SAS/documents/technical/education/ en/solve-data-science-skills-shortage-uk-113039.pdf. Deeg, Nils and Tania Pierotic. 2024. “Digital Policy Alert Digital Digest: South Africa (2024 Edition).” December. https:// digitalpolicyalert.org/ digest/dpa-digital-digest-south-africa. den Besten, Matthijs L. 2020. “Crunchbase Research: Monitoring Entrepreneurship Research in the Age of Big Data.” SSRN, November 3. https://ssrn.com/abstract=3724395. Elsner, Mark, Grace Atkinson and Saadia Zahidi. 2025. The Global Risks Report 2025. Cologny, Switzerland: World Economic Forum. January. www.weforum. org/publications/global-risks-report-2025/. Élysée. 2025. “The Paris Charter on Artificial Intelligence in the Public Interest.” Office of the President of France, February 11. www.elysee.fr/en/emmanuel-macron/2025/02/11/theparis-charter-on-artificial-intelligence-in-the-public-interest. Fan, Qimiao, and Christine Zhenwei Qiang. 2024. “Tipping the scales: AI’s dual impact on developing nations.” Digital Transformation (World Bank blog), June 3. https:// blogs.worldbank.org/en/digital-development/tippingthe-scales--ai-s-dual-impact-on-developing-nations. 50 CIGI Paper No. 323 — June 2025 • Douglas Lippoldt Filgueiras, Fernando and Tainá Aguiar Junquilho. 2023. “The Brazilian (Non)perspective on national strategy for artificial intelligence.” Discover Artificial Intelligence 3 (7). https://doi.org/10.1007/s44163-023-00052-w. Gao, Xueyuan and Hua Feng. 2023. “AI-Driven Productivity Gains: Artificial Intelligence and Firm Productivity.” Sustainability 15 (11): 8934. https://doi.org/10.3390/su15118934. G7 Italia. 2024. “Ministerial Meeting on Technology and Digital: G7 Joint Statement.” October 15. www.soumu.go.jp/ hiroshimaaiprocess/pdf/document08_en.pdf. Geradin, Damien. 2017. “Ensuring sound regulatory processes: For a principled approach.” TILEC Discussion Paper No. 2017-030. July. http://ssrn.com/abstract=3011188. Giardini, Tommaso, Luca Scheiwiler and Maria Buza. 2024. “Digital Policy Alert Digital Digest: Brazil.” December. https:// digitalpolicyalert.org/ digest/dpa-digital-digest-brazil. Government of Canada. 2024. “G7 Italia 2024– Addressing global challenges, fostering partnerships.” Foreign Ministers’ meeting communiqué, April 19. www.international.gc.ca/ world-monde/international_relations-relations_ internationales/g7/documents/2024-04-19parterships-partenariats.aspx?lang=eng. Government of Japan. 2024. “Documents of Achievement.” Hiroshima AI Process. www.soumu.go.jp/hiroshimaaiprocess/ en/documents.html.Achievement.” Hiroshima AI Process. www.soumu.go.jp/hiroshimaaiprocess/en/documents.html. GOV.UK. 2023. The Bletchley Declaration by Countries Attending the AI Safety Summit, \ 1–2 November 2023. Policy Paper. www.gov.uk/government/publications/ai-safety-summit2023-the-bletchley-declaration/the-bletchley-declaration-bycountries-attending-the-ai-safety-summit-1-2-november-2023. Heller, David and Dominik Asam. 2024. “Generative AI and Firmlevel Productivity: Evidence from Startup Funding Dynamics.” SSRN, August 21. https://ssrn.com/abstract=4877505. Hiter, Shelby. 2024. “150 Top AI Companies: Visionaries Driving the AI Revolution.” eWeek, October 1. www.eweek.com/ artificial-intelligence/ai-companies/. (Note: Updates to this article are posted periodically by eWeek.) IFAC-BIAC. 2018. Regulatory Divergence: Costs, Risks, Impacts. International Financial Sector Study at OECD and IFAC. https://bdi.eu/media/ themenfelder/konjunktur_und_finanzmaerkte/ downloads/Business_at_OECD-IFAC_Costs_ of_Regulatory_Divergence.pdf. Ipsos. 2023. Global Views on A.I. 2023. Global Advisor survey. July. www.ipsos.com/sites/ default/files/ct/news/documents/2023-07/ Ipsos Global AI 2023 Report-WEB_0.pdf. Kalash, S. Yash. 2024. “AI and the Digital Economy Framework Agreement: Preparing ASEAN for the Next Tech Wave.” Opinion, Centre for International Governance Innovation, November 11. www.cigionline.org/articles/ ai-and-the-digital-economy-framework-agreementpreparing-asean-for-the-next-tech-wave/. Korinek, Anton. 2024. “Economic Policy Challenges for the Age of AI.” National Bureau of Economic Research Working Paper 32980. September. www.nber.org/papers/w32980. Korinek, Anton and Joseph E. Stiglitz. 2021. “Artificial Intelligence, Globalization, and Strategies for Economic Development.” National Bureau of Economic Research Working Paper 28453. February. www.nber.org/papers/w28453. Krugman, Paul. 1995. Development, Geography and Economic Theory. Cambridge, MA: The MIT Press. Lippoldt, Douglas. 2024. AI Innovation Concentration and the Governance Challenge. CIGI Paper No. 292. Waterloo, ON: CIGI. www.cigionline.org/publications/ai-innovationconcentration-and-the-governance-challenge/. Lippoldt, Douglas C. and Mark F. Schultz. 2014. “Uncovering Trade Secrets – An Empirical Assessment of Economic Implications of Protection for Undisclosed Data” OECD Trade Policy Paper No. 167. Paris, France: OECD Publishing. August 11http://dx.doi.org/10.1787/5jxzl5w3j3s6-en Lund, Brady and Wang Ting. 2023. “Chatting about ChatGPT: How may AI and GPT impact academia and libraries?” Library Hi Tech News 40 (3): 26–29. https://doi.org/10.1108/LHTN-01-2023-0009. Mackowski, Martin, Pablo Carrillo, and Julia Jacobson. 2025. “Trump’s AI Policy Shift Promotes US Dominance and Deregulation.” Bloomberg Law, February 26. https:// news.bloomberglaw.com/us-law-week/trumps-ai-policyshift-promotes-us-dominance-and-deregulation. Mannuru, Nishith Reddy, Sakib Shahriar, Zoë A. Teel, Ting Wang, Brady D. Lund, Solomon Tijani, Chalermchai Oak Pohboon et al. 2023. “Artificial intelligence in developing countries: The impact of generative artificial intelligence (AI) technologies for development.” Information Development: 1–19. September. https://doi.org/10.1177/02666669231200628. Martins, Laís. 2025. “Brazil’s AI Law Faces Uncertain Future as Big Tech Warms to Trump.” Tech Policy Press, February 4. www.techpolicy.press/brazils-ai-lawfaces-uncertain-future-as-big-tech-warms-to-trump. Maslej, Nestor, Loredana Fattorini, Raymond Perrault, Vanessa Parli, Anka Reuel, Erik Brynjolfsson, John Etchemendy et al. 2024.