The Forum on AI Risk and Resilience (FAIRR)

Share this Chapter


While domestic measures are the required foundation for the governance of GenAI, its transnational nature makes international mechanisms a necessary corollary.1 Domestic governance cannot function if actors can operate beyond a state’s jurisdiction. GenAI tools cross borders with relative ease.2 An international regime is required to close gaps. A series of tailored mechanisms are most appropriate as governance arrangements work best when fitted to specific issues and challenges. GenAI presents risks at the nexus of geopolitical, transnational, and societal concerns. This memo focuses on a subset of three interconnected aims: 

  1. Preventing non-state malign GenAI use for nefarious ends (e.g., criminal activities or acts of terrorism); 
  2. Mitigating the most consequential, injurious GenAI impacts on society (e.g., illegitimate discriminatory impacts due to system bias); and 
  3. Managing GenAI use that infringes on other states’ sovereignty (e.g., foreign malign influence operations or the use of AI tools in cyber surveillance).  

SCSP proposes a new multilateral and multi-stakeholder Forum on AI Risk and Resilience (FAIRR). That new regime must function within the realities of today’s international landscape – deepening geopolitical tensions and limited trust. The world has met similar challenges before. The Cold War superpowers recognized shared challenges and created tools to manage them.3 More recently, in confronting transnational terrorism, climate change, and financial crisis, nations have sought to avoid ungoverned spaces as fonts of instability.4 Therefore, FAIRR is the path forward as a new international entity scoped in mission and inclusive in nature.


FAIRR’s work would center on three interconnected governance needs. First, nation-states share broadly common ground in preventing misuse and malign use of GenAI tools by non-state actors. Illicit uses of GenAI tools that contribute to harms from economy-sapping fraud5 to bio-terrorism6 are detrimental to all. Second, while the contours might differ state-by-state, countries generally agree on the need to prevent harms such as discrimination.7 Governance alignment to prevent these harms from creeping across borders and negatively impacting populations would constitute a likely point of at least partial agreement. Third, states require a forum to discuss and deconflict instances where GenAI tools can cross boundaries and impact the sovereign activities of other states. For instance, the world will require a forum for dialogue to develop norms around legitimate GenAI-enhanced speech in an era of both growing misinformation and transnational censorship, potentially including around elections. Finally, pursuing all three sets of issues within one entity would help address the blurred lines between the three spaces. Particularly in the digital realm, states have deployed non-state actors for state ends.8  


FAIRR would not constitute a new formal international agency, with international legal personality grounded in a treaty and creating binding rules.9 Hurdles from treaty ratification10 to skepticism of international institutions11 preclude that path. Instead, FAIRR would serve as a forum to promulgate alignment based on soft law.12 FAIRR would convene relevant stakeholders — including national officials, regulators, relevant private sector companies, academia, and civil society — to work toward interoperable standards and rules that domestic regulators can independently implement. 


Establishing FAIRR under the auspices of the G20 would both provide a legitimate foundation and establish a basis for its founding members. As with the Financial Stability Board (FSB),13 FAIRR would benefit from being sufficiently inclusive to cover enough of the global economy to drive both member and non-member adherence to the established rules. Such an impetus would support seeking PRC participation in order to work toward globally comprehensive and consistent practices. Less ambitiously, even if the members became deadlocked, the PRC’s participation in FAIRR would help avoid forum shopping14 that could lead to the PRC rallying an international grouping under alternative rules inimical to U.S. values and interests.15

Furthermore, while drawing on a G20 foundation, FAIRR should voluntarily go beyond that core G20 membership in two respects. First, it should include as full original members any state with a headquarters of a GenAI actor over a certain compute threshold, such as the United Arab Emirates (UAE).16 Such a move would help avoid nations with GenAI capabilities looking to caucus states to build alternative regimes. Second, as GenAI is first and foremost a private sector-driven advancement,17 FAIRR should be multi-stakeholder in composition, including major private entities with GenAI capabilities over a certain compute threshold as full voting members. A weighted voting system18 would help ensure national governments remained dominant over private entities in order to maintain the primacy of public interests. Additionally, an advisory committee(s) of academic and civil society experts could help balance private sector interests at the table. 


Full members, original members (state and non-state), plus states that grow to meet the criteria of being home to an above-threshold model, would set relevant standards and rules. Members would participate in a double-weighted voting system to capture both state and non-state equities and disparate capabilities. Under this system, the total voting share of states would equal twice the voting shares of non-state members. Simultaneously, to distinguish among state capabilities, the total share of votes for states would be proportional to the number of qualifying private GenAI entities in their jurisdiction (with a minimum of one vote per state).  

Similar to the FSB or the Financial Action Task Force (FATF),19 a peer review process of domestic regulators would provide political pressure to abide by commonly set terms.20 Noncompliant members would confront market ramifications based on their recognized higher risk, such as business withdrawals or chilled investment in response.21 


Resourcing must take place in a manner that maintains FAIRR’s legitimacy. Perceptions of private funding leading to corporate interest capture would undercut the entire initiative.22 Consequently, funding for the secretariat and operating costs should fall primarily (majority) with national contributions, likewise differentiated according to vote share. Private companies would contribute a minority share of the budget with proportions divided based on market capitalization.


  1. While still in its early stages, there is a growing body of scholarship exploring international AI governance and historical precedents. See Michael Veale, et al., AI and Global Governance: Modalities, Rationales, Tensions, Annual Review of Law and Social Science (2023); Ian Stewart, Why the IAEA Model May Not be Best for Regulating Artificial Intelligence, Bulletin of the Atomic Scientists (2023); Peter Cihon, et al., Should Artificial Intelligence Governance be Centralised? Design Lessons from History, Proceedings of Proceedings of the 2020 AAAI/ACM Conference on AI, Ethics, and Society (2020).
  2. For an example of concerns of how AI will proliferate across jurisdictions, see Brian Nussbaum, Offshore: The Coming Global Archipelago of Corrosive AI, Lawfare (2023). 
  3. U.S.-Soviet cooperation covered a range of issues from bilaterally collaborating on research in the health sciences to establishing the IAEA to curb the proliferation of nuclear weapons. See Bernard Gwertzman, U.S. and the Soviet to Pool Research in 3 Health Areas, New York Times (1972); Bertrand Goldschmidt, The Origins of the International Atomic Energy Agency, IAEA Bulletin at 15-16 (1977).
  4. See Bruce Jones, et al., Power and Responsibility: Building International Order in an Era of Transnational Threats, Brookings Institution Press (2009). 
  5. Sam Sabin, Generative AI is Making Voice Scams Easier to Believe, Axios Codebook (2023).
  6. See John T. O’Brien & Cassidy Nelson, Assessing the Risks Posed by the Convergence of Artificial Intelligence and Biotechnology, Health Security (2020).
  7. While states can vary in their interpretation, wide acceptance of the International Convention on the Elimination of All Forms of Racial Discrimination speaks to a broad acceptance of this goal. See International Convention on the Elimination of All Forms of Racial Discrimination,1969 (EIF), 660 UNTS 195.
  8. Erica D. Lonergan, Cyber Proxies in the Ukraine Conflict: Implications for International Norms, Council on Foreign Relations (2022). 
  9. See Anne Burnett, International Organizations, American Society of International Law at 3-4 (2015).
  10. For instance, the U.S. domestic political environment remains skeptical of ratifying treaties. See Anya Wahal, On International Treaties, the United States Refuses to Play Ball, Council on Foreign Relations (2022). 
  11. See Stewart Patrick, The Sovereignty Wars: Reconciling America with the World, Brookings Institution Press at 227-32 (2018); Julian Ku & John Yoo, Taming Globalization: International Law, the U.S. Constitution, and the New World Order, Oxford University Press at 42-47 (2012).
  12. See Gary Marchant, “Soft Law” Governance of Artificial Intelligence, UCLA AI Pulse (2019); Kenneth W. Abbott & Duncan Snidal, Hard and Soft Law in International Governance, International Organization (2000). 
  13. See About the FSB, Financial Stability Board (2020); Stavros Gadinis, The Financial Stability Board: The New Politics of International Financial Regulation, Texas International Law Journal at 165 (2013). Robert Fay at the Centre for International Governance (CIGI) has led the way in exploring adapting the FSB structure to the broader digital domain. See Robert Fay, Global Governance of Data and Digital Technologies: A Framework for Peaceful Cooperation, Centre for International Governance Innovation (2022); Robert Fay, Digital Platforms Require a Global Governance Framework, Centre for International Governance Innovation (2019). 
  14. The concept of “forum shopping” in international scenarios builds on the concept of the domestic U.S. legal practice “of pursuing a claim subject to concurrent jurisdiction in the court that will treat the claim most favorably.” Forum Shopping, Cornell Legal Information Institute (2022). On forum shopping in the context of international AI governance, see Peter Cihon, et al., Fragmentation and the Future: Investigating Architectures for International AI Governance, Global Policy at 550 (2020).
  15. As the PRC’s Regional Comprehensive Economic Partnership (RCEP) initiative illustrates, the PRC is more than capable of seizing open ground to propose its own standards and norms. See Michael Sutherland, Regional Comprehensive Economic Partnership (RCEP), Congressional Research Service (2020). 
  16. In May 2023, the UAE unveiled a 40 billion parameter LLM, the Falcon 40B. UAE’s First LLM is Open Source and Tops Hugging Face Leaderboard, Wired (2023).
  17. See Harnessing the New Geometry of Innovation, Special Competitive Studies Project at 22-29 (2022).
  18. While the FSB’s Plenary operates on consensus, it offers a model in which countries receive different degrees of representation based on “the size of the national economy, financial market activity and national financial stability arrangements of the corresponding Member jurisdiction.” Charter of the Financial Stability Board, Article 11 (2012). See also Diego Lombardi, The Governance of the Financial Stability Board, Brookings Institution at 10-11 (2011).
  19. See What We Do, Financial Action Task Force (last accessed 2023).
  20. See Peer Reviews, Financial Stability Board (2021); Mutual Evaluations, Financial Action Task Force (last accessed 2023). See also Stavros Gadinis, The Financial Stability Board: The New Politics of International Financial Regulation, Texas International Law Journal at 160 (2013). 
  21. As a comparable example, see the impact of FATF gray listing on foreign investment in Pakistan. Purvaja Modak, FATF’s Scrutiny and What Non-Compliance Means, Centre for Public Policy Research (2021). 
  22. Regulatory Capture, Oxford Reference (last accessed 2023).