At the 80th session of the United Nations (UN) General Assembly in New York last month, the UN Member States formally launched two new initiatives: the Global Dialogue on AI Governance and the Independent International Scientific Panel on AI. Both derive from the Global Digital Compact, and they were legitimized in resolution A/RES/79/325, officially adopted on August 26. Prior to their inclusion in the Global Digital Compact, both initiatives were among the recommendations reported in September 2024 by the High-level Advisory Body on Artificial Intelligence, initially proposed in 2020 as part of the United Nations Secretary-General’s Roadmap for Digital Cooperation (A/74/821).
The Global Dialogue on AI Governance will “provide an inclusive platform within the United Nations for states and stakeholders to discuss the critical issues concerning AI facing humanity today.” It will promote interoperability between different strands of governance and encourage open innovation. By complementing existing international initiatives (such as the G7’s Hiroshima AI Process), it aims to “provide an inclusive, stable home for coordinating AI governance that will help build safe, secure and trustworthy AI systems.” The focus on inclusion of the 193 member states has been welcomed by the World Economic Forum, which notes that “as the capabilities and deployment of AI continue to accelerate worldwide, 118 countries are not party to any significant international AI governance initiative,” and “just seven countries – all from the developed world – are parties to all the current significant global AI governance initiatives.”
The Independent International Scientific Panel on AI will “serve as a crucial bridge between cutting-edge AI research and policymaking. By providing rigorous, independent scientific assessments, it will help the international community to anticipate emerging challenges and make informed decisions about how we govern this transformative technology.” It will be comprised of 40 global experts tasked with producing an annual scientific report on the risks, opportunities, and impact of AI. An open call for applicants to serve in their personal capacity for a three-year term closes on October 31. It will be led by two co-chairs (one from a developed country, the other from a developing country). Members will be appointed on the basis of their outstanding expertise in AI and related fields, an interdisciplinary perspective, and geographical and gender balance. No more than two candidates from the same nationality or affiliation will be selected.
The UN’s new AI governance architecture has been proclaimed a “symbolic triumph.” It is ambitious and represents the world’s most globally inclusive approach to governing AI. Critics and supporters alike have called the panel an IPCC for AI. But what can it achieve?
First, international representative-based institutions focused upon reaching consensus are ill-suited to keep up with rapidly-changing technologies, their capabilities, and their potential. Neither are they necessarily capable of responding effectively, and in a timely manner, to unpredictable crises. It seems difficult enough to achieve cooperation, even when outcomes are known and predictable; it’s unlikely that they will prove any more effective in response to a truly unanticipated “black-swan” AI event.
Second, even if coordinated action was possible, the UN lacks the means of enforcing any resolutions. Cooperation is voluntary, with any actions being the responsibility of individual member states. If any major AI-developing nation state (e.g., the US or China) pulls back from participating or complying, any resolutions will be effectively impotent.
Third, funding this ambitious initiative is fraught with difficulties. All UN activities are contingent upon funding by member states, most notably the United States. Given the recent executive order confirming plans to reassess all international organizations, it is unclear how financially sustainable—or globally representative or independent—the new initiatives will actually be. While no payments will be made to the scientific panel members, producing reports isn’t cheap. Moreover, the ability to obtain the degree of representation required from scientific experts in the global south may be difficult if they are already resource-constrained in their existing positions. This threatens the quality of both the research and any policy developed from it.
Finally, it begs the question of whether the purpose of the initiatives is to foster the best possible world-leading research on AI governance for safety, transparency, and trustworthiness, or to foster research and promulgation of rules meeting a much less focused distributional objective regarding the benefits from using AI to conform with meeting a wide array of UN social development goals. It seems somewhat naive to expect that rules governing a means of production can also deliver equitable outcomes from the consumption of its outputs.
So while the UN may deserve some credit for achieving imperfect international cooperation (better than no cooperation) on AI, the jury is out on its potential impact or influence.