Using artificial intelligence in venture fund management: legal risks and liability

International Law, Business and Political Science Journal ISSN-L 3235-9799 E-ISSN 3235-9799 IF(Impact Factor) 13.24 https://journallaw.totalh.net/ Volume: 11. Issue 12 November 2025 1 Using artificial intelligence in venture fund management: legal risks and liability Kayumova Asalya Sultonmurodovna teacher of the Department of Civil Law at Tashkent State University of Law ORCID: 0009-0006-7609-6223 e-mail: [email protected] Abstract: Artificial intelligence (AI) has begun to transform the venture capital (VC) industry by automating analytical tasks, supporting due diligence, and improving decision-making efficiency in investment selection and portfolio management. AI technologies assist venture fund managers in identifying high-growth startups, predicting market behavior, evaluating intellectual property, and optimizing exit strategies. However, the use of AI also generates new categories of legal risks, including data governance failures, algorithmic bias, lack of transparency, cybersecurity vulnerabilities, and ambiguity in the allocation of liability between fund managers and AI developers. This article examines the legal implications of adopting AI in venture fund management through a comparative legal-analytical approach. Using international regulatory sources from ESMA, FCA, IOSCO, NIST, and academic studies, it evaluates challenges related to fiduciary duties, accountability, investor protection, and model governance. Particular focus is placed on the evolving legal framework in Uzbekistan, where venture regulation is still developing and lacks explicit AI governance norms. The research concludes that although AI creates significant operational benefits, venture fund managers remain fully responsible for investment outcomes, and regulatory reforms are required to ensure investor protection and mitigate legal uncertainty. Keywords: artificial intelligence, venture capital, venture funds, investment management, fiduciary duty, legal risks, liability, regulatory compliance. International Law, Business and Political Science Journal ISSN-L 3235-9799 E-ISSN 3235-9799 IF(Impact Factor) 13.24 https://journallaw.totalh.net/ Volume: 11. Issue 12 November 2025 2 Introduction Venture capital plays a central role in funding early-stage companies with high growth potential, allowing innovative startups to commercialize ideas under conditions of uncertainty and market volatility. A venture fund typically pools capital from investors, managed by a management company or general partner (GP), which assumes fiduciary responsibility for investment decisions. The introduction of artificial intelligence (AI) into venture fund operations marks a paradigm shift in how investment decisions are made. Unlike traditional methods that depend on expert judgment, network-based sourcing, and manual qualitative evaluation, AI enables systematic processing of large and unstructured data, enhancing accuracy and reducing time spent on due diligence and monitoring. Machine learning models can assess market trends, estimate startup scalability, analyze founders’ track records, and predict failure or success with measurable precision [1, p. 16]. The key participants in the application of AI within venture capital fund management include: 1. Investors (Limited Partners), who provide capital and remain interested in the proper and lawful management of the fund; 2. The management company (General Partner), which makes strategic AIsupported investment decisions and is responsible for compliance and risk control; 3. AI technology providers, who supply analytical platforms, predictive models, or machine-learning tools integrated into the investment process. While AI systems may significantly influence the evaluation of portfolio companies, the management company continues to bear ultimate legal responsibility for investment outcomes [1, p.17]. When incorporating artificial intelligence into the internal processes of a venture capital fund, particular attention is given to the alignment of AI-driven decision-making International Law, Business and Political Science Journal ISSN-L 3235-9799 E-ISSN 3235-9799 IF(Impact Factor) 13.24 https://journallaw.totalh.net/ Volume: 11. Issue 12 November 2025 3 with regulatory requirements, data governance frameworks, and liability allocation. At the initial stage, the fund may face challenges related to legal risk allocation, data protection, model reliability, transparency, and tax treatment of digital services. When designing AI-supported investment activities, the management company or fund participant typically pursues the following main objectives: I. Limitation of liability – although AI systems assist investment analysis, they cannot replace fiduciary obligations. Thus, one of the primary objectives is to ensure that the responsibilities of the management company are clearly defined and that limited partners retain their limited liability, even when investment decisions are partially based on AI models. Clear agreement on liability allocation prevents the misinterpretation of AI output as a substitute for human professional judgment [2, p. 7]. II. Avoiding double taxation and minimizing tax exposure related to digital services – the use of paid AI platforms, cloud-based analytics, or outsourced model development may create additional tax burdens. As with traditional fund structuring, minimising double taxation on returns generated through AI-influenced decisionmaking remains important. In addition, managing service fees for AI platforms – especially when acquired cross-border – requires accurate VAT treatment and contractual planning. III. Optimality and suitability for all investors – the deployment of AI tools should not create discriminatory investment conditions or introduce opacity for investors unfamiliar with algorithmic processes. To maintain investor confidence, fund operations must remain transparent and understandable. This also allows institutional and foreign investors to participate without regulatory disadvantages arising from AIbased practices. IV. Targeting appropriate investment segments – AI systems allow venture funds to specialise more precisely, for example, in technology-intensive or scientific fields, by processing sector-specific data and providing market-relevant forecasting. International Law, Business and Political Science Journal ISSN-L 3235-9799 E-ISSN 3235-9799 IF(Impact Factor) 13.24 https://journallaw.totalh.net/ Volume: 11. Issue 12 November 2025 4 This supports targeted fundraising and assists investors in evaluating sectoral risks. The focus on specific technological domains, such as fintech or biotechnology, enables the development of dedicated AI models trained on relevant datasets [3, p. 4]. V. Ease of management and regulatory compliance — AI integration must be compatible with the fund’s governance structure and compliance framework. Efficient management requires that AI-based decisions remain explainable, auditable, and aligned with internal policies. In case of procedural inconsistency, the fund must prioritize compliance and fiduciary duties over technical efficiency. Therefore, appropriate contractual, technological, and supervisory mechanisms must be implemented to ensure that AI does not compromise regulatory adherence or internal control systems [4, p. 5]. We have to mention that, the use of AI in VC reflects broader global trends in asset management. Studies indicate that more than two-thirds of fund managers are experimenting with AI-based tools in risk management, market mapping, and predictive analytics [2, p. 4]. In VC, these tools add value in early-stage investment where informational asymmetry is high and traditional financial statements are absent. In Uzbekistan, legal regulation of venture capital is in its early stages. While venture fund structures are being introduced, specific norms governing AI-assisted management have not yet been developed. Therefore, core obligations derive from general civil-law duties and contractual arrangements within management agreements. This study builds upon and parallels structural analyses of venture funds previously undertaken and early reflections on AI usage, but expands the scope to include a comparative analysis of international standards and their relevance for Uzbekistan. Materials and Methods of Research The research employs a qualitative analytical methodology based on: Doctrinal legal analysis – examination of legal concepts of fiduciary duty, International Law, Business and Political Science Journal ISSN-L 3235-9799 E-ISSN 3235-9799 IF(Impact Factor) 13.24 https://journallaw.totalh.net/ Volume: 11. Issue 12 November 2025 5 liability, and investment governance in relation to the use of AI in venture fund management. Comparative legal analysis – study of approaches adopted by the EU, UK, U.S., Asia, and international regulatory bodies (ESMA, FCA, IOSCO, NIST) to identify rules applicable to AI-supported investment decision-making. This comparison is contrasted with the current legal landscape in Uzbekistan, where specific AI governance remains undeveloped. Document analysis – study of technical and policy documents including: 1. ESMA reports on AI in investment funds (2025) [4, p. 2] 2. ESMA statement on AI and investment services (2024) [5, p. 2] 3. FCA AI Update (2024) [5, p. 4] 4. IOSCO guidance on AI risks (2024) [6, p. 7] 5. NIST AI Risk Management Framework (2023) [7, p. 6] 6. Academic works on corporate governance and AI [1, p. 5] 7. Industry commentary on VC and AI adoption [2, p. 4; 11, p. 6] Case-oriented review of VC activity – Examination of how leading venture funds apply AI to due diligence, portfolio monitoring, and exit strategies based on publicly available analyses. Inductive reasoning - used to synthesize conclusions regarding legal risks and formulate recommendations for Uzbekistan. The methodology enables a holistic understanding of how AI technologies interact with legal structures governing venture fund management, revealing both opportunities and risk points. Research results Artificial intelligence (AI) is gradually becoming an integral part of the venture capital (VC) industry. Unlike traditional financial institutions in which standardized instruments and historical financial data dominate decision-making, venture funds International Law, Business and Political Science Journal ISSN-L 3235-9799 E-ISSN 3235-9799 IF(Impact Factor) 13.24 https://journallaw.totalh.net/ Volume: 11. Issue 12 November 2025 6 invest in innovative enterprises whose business models and technologies are often experimental. Thus, reliance on AI tools offers an opportunity to reduce informational asymmetry, accelerate evaluation, improve portfolio monitoring, and mitigate risks associated with early-stage uncertainty. However, this integration also generates significant legal risks and raises questions about the allocation of responsibility between participants of a venture fund. AI affects the entire investment cycle — from deal sourcing to due diligence, valuation, and exit strategy. Modern AI tools analyze founders’ backgrounds, technological feasibility, market development, and sentiment indicators through machine learning, Big Data processing, and natural language analysis. According to industry research, more than half of global venture funds now use some form of automated decision-making or predictive analytics to identify potential portfolio companies, evaluate intellectual property strength, and model exit prospects [2, p. 4]. For example, algorithms developed in the U.S. and Asian markets assess thousands of early-stage companies based on patterns recognizable only through computational analysis [11, p. 6]. Nevertheless, while AI can increase efficiency and accuracy, it introduces new legal and operational risks. These risks can be categorized into five major groups: fiduciary-duty risks, data governance risks, model risks, discrimination risks, and liability-allocation risks, like: 1. Fiduciary-duty risks and responsibility of the management company. In most jurisdictions, including the European Union and Uzbekistan, the general partner (GP) or management company of a venture fund bears fiduciary duties to limited partners (LPs). These duties typically include the duty of care, the duty of loyalty, and the duty to act in the best interests of investors. The use of AI does not abolish or diminish these obligations. As ESMA notes, firms employing AI in investment processes remain fully International Law, Business and Political Science Journal ISSN-L 3235-9799 E-ISSN 3235-9799 IF(Impact Factor) 13.24 https://journallaw.totalh.net/ Volume: 11. Issue 12 November 2025 7 responsible for regulatory obligations, including suitability evaluation, governance, and consumer protection [4, p. 8]. Even if an algorithm suggests a particular investment, a GP must still review it, verify its reasonability, and act prudently. Similar positions are stated by the UK’s Financial Conduct Authority (FCA), which emphasizes that the use of automated systems does not justify poor decision-making or rule violations [5, p. 6]. From a civil-law perspective, AI tools may be qualified as auxiliary instruments assisting professional judgment; thus, their use does not shift statutory duties to a third party. Should losses occur due to reliance on an AI system, the management company may still be deemed negligent unless it can demonstrate proper oversight, testing, and justification. In Uzbekistan, the legal framework for venture funds is still forming, and there is no explicit regulation of AI applications in investment management. However, fiduciary-duty principles, derived from general civil liability and contractual arrangements, indicate that responsibility remains with the management company, regardless of algorithmic participation. Thus, Uzbekistan’s regulatory conditions currently mirror international practice in assigning baseline accountability to managers. 2. Data governance and confidentiality risks. AI systems rely on large datasets, including financial data, proprietary information, and personal data. Poor governance of such data may result in privacy breaches, intellectual property loss, or violation of confidentiality agreements. ESMA has highlighted that AI-based investment services must ensure data security and integrity while maintaining full compliance with data-protection legislation [4, p. 10]. Similarly, IOSCO states that data quality, storage controls, anonymization standards, and data provenance are essential to prevent misuse or unlawful processing [6, p. 3]. This presents particular challenges in cross-border VC operations, especially when investing in startups that develop sensitive technologies. Data transfers may be International Law, Business and Political Science Journal ISSN-L 3235-9799 E-ISSN 3235-9799 IF(Impact Factor) 13.24 https://journallaw.totalh.net/ Volume: 11. Issue 12 November 2025 8 subject to foreign-jurisdiction approvals. In Uzbekistan, although data protection laws exist, they do not yet address AI-specific risk categories, leaving uncertainty regarding compliance in international VC transactions. Failure to ensure high-quality, legally compliant datasets can also distort investment decisions by producing inaccurate or biased model outcomes. 3. Risks of algorithmic bias and discrimination. Machine-learning models used in venture fund management may suffer from hidden biases linked to training information or model structure. These biases can result in discriminatory patterns in investment selection, e.g.: favoring companies from certain geographic regions; preferring founders with specific backgrounds or demographic traits; ignoring innovative but unconventional business models. Such patterns reproduce inequalities embedded in historical data and contradict fair-treatment principles. Although no widely ratified antidiscrimination rules exist specifically for VC operations, many national laws include equality principles that may indirectly apply. Moreover, biased investment strategies may cause:  legal claims from disadvantaged parties;  reputational harm;  reduced market competitiveness. ESMA and NIST emphasize the need for continuous model evaluation, explainability, and mitigation procedures against discriminatory outcomes [4, p. 14; 7, p. 6]. In Uzbekistan, while AI-specific anti-bias requirements remain undeveloped, general nondiscrimination principles apply and may serve as grounds for claims where bias results in economic harm. 4. Model risk and explainability challenges. AI models can malfunction due to incorrect parameters, poor training data, or unexpected market developments. Such International Law, Business and Political Science Journal ISSN-L 3235-9799 E-ISSN 3235-9799 IF(Impact Factor) 13.24 https://journallaw.totalh.net/ Volume: 11. Issue 12 November 2025 9 failures may lead to inaccurate valuations, flawed modeling, or misidentification of risks. The challenge of explainability — the difficulty of understanding how a model arrives at a given output — limits the ability of fund managers to detect errors or justify decisions to investors. ESMA warns that opacity in AI decision-making complicates supervisory and audit functions [4, p. 11]. In turn, FCA requires that automated systems used in investment processes should be documented and subject to human oversight [5, p. 7]. For venture capital, where traditional due diligence is replaced partly by algorithmic evaluation, depth of explainability becomes crucial. Failure to explain reasoning may result in the misinterpretation of risk profiles and unclear justification for investment choices, creating grounds for investor disputes. 5. Allocation of liability in AI-driven investment decisions. A central legal problem in AI-assisted VC management concerns how to allocate responsibility when AI contributes to poor decisions. I. Potentially responsible actors include: II. The management company (GP) III. AI developers or service providers IV. Data suppliers V. Portfolio companies (in cases of misrepresentation) Even when AI developers contribute to faulty decisions, international regulatory frameworks uphold that fund managers remain primarily accountable. ESMA and FCA expressly state that outsourcing or automating decision-making does not transfer regulatory responsibility to technology providers [4, p. 8; 5, p. 9]. In civil-law jurisdictions, including Uzbekistan, liability may also be assigned to the management company as the principal organizer of investment activities. Contractual mechanisms (indemnities, warranties) may provide internal recourse but do