Exclusive: Biden administration douses UN's AI governance aspirations

The Biden administration has poured cold water on United Nations Secretary-General António Guterres’ aspiration of establishing new global institutions to govern the use of artificial intelligence, saying the world body should instead focus on strengthening existing international entities that promote lifesaving initiatives, develop ethical norms, and erect guardrails around the misuse of the transformative technology. “It would be premature to call for establishing new U.N. governance mechanisms without a clearer understanding and fulsome consensus on where may be gaps in the ability of existing U.N. agencies to address AI,” according to a confidential United States paper distributed to foreign governments at the U.N. that was seen by Devex. “It is our view that any proposals for new processes, panels, funds, partnerships, and/or mechanisms are premature.” The blunt U.S. message deals a severe blow to the U.N. chief’s efforts to place the world body at the center of global efforts to harness AI to combat poverty and inequality while erecting safeguards around a technology that is advancing at a disorienting pace and which holds the power to fundamentally alter the way societies grapple with everything from infectious diseases to economic development and war. The U.S. move comes as U.N. delegations in New York prepare to resume negotiations on May 2 on the terms of a Global Digital Compact, or GDC, an international pact setting the digital rules of the road. It is expected to be endorsed by world leaders at the U.N. Summit of the Future in September. “We recognize the central role of the UN in supporting AI governance,” an early draft, dated April 1, states. “We have a unique opportunity, through this Compact, to establish the foundations of governance of AI.” The release of the U.S. paper comes in response to a series of recommendations by a panel of some 38 experts appointed by Guterres in December to advance the U.N.’s role in AI governance. The so-called High-Level Advisory Body on AI, or HLAB on AI, published an interim report, titled “Governing AI for Humanity,” arguing for a central role for the U.N. in the global governance of artificial intelligence. Its views are believed to align with those of the U.N. leadership. “This technology cries out for governance,” the report states. “The United Nations lies at the heart of the rules-based international order.” The advisory group proposed, during a March 25 briefing to member states, the creation of a series of entities, including a scientific panel on AI that would report every six months on the latest advances in AI, and provide U.N. members with early warning on potential risks of the fast-evolving technology, according to a copy of the presentation reviewed by Devex. It also recommended the establishment of a high-level contact group that would meet annually to discuss AI governance; a non-U.N. technical forum for harmonizing AI standards; and an international public-private partnership that promotes capacity building and technology sharing in the developing world. It called for the establishment of a global fund for AI with $10 billion in start-up money, to ensure artificial intelligence serves the public over commercial interests. “A coordinated and institutionally networked approach with the UN at the heart of it might be the best way to realize digital cooperation and digital solidarity on AI,” the presentation states. “The UN is a unique universal platform: No other institution has the inclusiveness, the legitimacy, the global convening power.” HLAB also proposed setting up a “small, agile” AI office within the U.N. secretariat to lay the groundwork for the broader reforms outlined in its presentation. HLAB conceded it would be premature to move ahead at the moment with a controversial informal U.N. proposal — floated in the summer of 2023 by Guterres — to set up an international entity modeled on the International Atomic Energy Agency to monitor and verify the international use of AI for peaceful purposes. But it said that “there should be some ‘red lines’ regarding military use of AI technologies and the autonomous targeting and harming of human beings by machines. But that could be addressed by delegations at the Convention on Certain Conventional Weapons in Geneva.” Guterres recently raised concern about the introduction of AI into warfare, saying he is “deeply troubled by reports that the Israeli military’s bombing campaign in Gaza includes artificial intelligence as a tool to identify targets, particularly in densely populated residential areas, resulting in a high level of civilian casualties.” “I have worried for many years of the dangers of weaponization [of] artificial intelligence and reducing the essential role of human agency,” he added. “AI should be a force for good to benefit the world, not to contribute to waging war on an industrial level, blurring accountability.” Turf war The latest U.N. reform initiatives on digital technology have roots in a 2020 request by the U.N. membership to revitalize a faltering multilateral system that appeared paralyzed in the face of the COVID-19 pandemic, which exposed the deep divisions among nations. The following year, Guterres outlined a major reform initiative, known as the Common Agenda, to restore vigor into a distressed multilateral system hamstrung by big power divisions. It includes the development of GDC to “set out principles, objectives, and actions to advance an open free, secure and human-centered digital future, one that is anchored in universal human rights and that enables the attainment of the Sustainable Development Goals.” The compact also addresses AI governance issues. In 2022, Guterres appointed a former Indian diplomat, Amandeep Singh Gill, as his new tech envoy, to lead U.N. efforts on the digital compact, and to sell the U.N. membership on a critical U.N. role in AI governance. In October 2023, Guterres established the HLAB advisory board to address “risks, opportunities and international governance of artificial intelligence.” The negotiations have opened a turf war between the U.N. tech envoy, Gill, who is championing a central role for the world body in AI governance, and U.N. agencies — like the International Telecommunication Union, which hosts an annual forum on AI, and UNESCO, which developed ethical guidelines for AI. They argue they already do what the tech envoy is proposing, according to U.N.-based diplomatic sources. In recent months, the Biden administration has sought to set parameters over the U.N.’s role in artificial intelligence, spearheading negotiations on the adoption in March of a U.S.-drafted U.N. General Assembly Resolution on AI. The landmark resolution, which was co-sponsored by more than 120 countries, vows to achieve “global consensus on safe, secure and trustworthy artificial intelligence systems” that promotes progress on the U.N.’s sustainable development goals. It also pledges to bridge the technical divide between and within countries, ensure AI developments observe U.N. human rights standards, and help spur private sector innovation. The resolution, according to the U.S. Department of State, “will serve as a foundation for multilateral AI efforts and existing and future UN initiatives.” While the U.S. acknowledges “the unique role for the U.N. system in achieving consensus” on AI policies, “it does not conclude that the UN system should be at the center of multilateral AI work by member states or stakeholders,” according to the internal U.S. paper. Instead, “the existing body of work already underway in these organizations in the domain of AI should also be taken into account, built upon, and strengthened as needed, rather than duplicated or undermined,” the U.S. paper added. “The UNGA resolution should not be misconstrued as identifying gaps and calling for new functions to be performed by new UN agencies.” The U.S., meanwhile, picked apart HLAB’s recommendations, noting, for instance, that the proposed international panel “may find it difficult to achieve ‘consensus’ on the technical aspects of AI, making such a proposal potentially counterproductive.” The role of the U.N. as a forum for debate on AI governance could potentially duplicate existing fora, like the World Summit on the Information Society, or WSIS, which was established in 2003 to promote the use of communications technology to advance the cause of development; and the Internet Governance Forum, which draws together governments, civil society organizations, and the private sector, as well as academics, technical experts, according to the U.S. and other delegations. “We should explore how to better support existing U.N. mechanisms and entities,” according to the internal U.S. paper. “For example, the Food and Agriculture Organization (FAO) would be well-placed to consider how AI can be used to address food security, while the World Meteorological Organization (WMO) could consider the impact of AI on weather and climate forecasting.” “Multiple UN entities are exploring ways to support capacity building and leverage AI to advance the SDGs. The UN Development Programme (UNDP); the UN Educational, Scientific, and Cultural Organization (UNESCO); the International Telecommunications Union,” it adds. “It is critical to hear from all stakeholders, including the private sector, the scientific and technical communities, and civil society, in international AI conversations, and those conversations and ultimate actions should not necessarily be centralized in the U.N. system,” according to the U.S. paper. A geopolitical minefield The draft declaration on the digital compact currently under discussion at U.N. headquarters includes some of HLAB’s proposals, including the creation of a scientific AI panel, as well as a Global Fund for AI and Emerging Technologies for Sustainable Development, but with a much smaller startup funding target of $100 million — considerably less than the initial $10 billion sought — “financed by voluntary contributions from public, private and philanthropic sources.” It places a greater emphasis on strengthening the role of existing international standards, governance entities, and debate forums, including the International Telecommunications Union, the World Summit on the Information Society, and the Internet Governance Forum. It also calls for the U.N. General Assembly to establish a dedicated U.N. office for coordinating digital and emerging technologies. “We support the proposed Global Digital Compact’s goals of advancing sustainable development and closing digital divides,” said U.S. Ambassador Chris Lu, who oversees U.S. negotiations on GDC. “We are actively reviewing the proposals on AI governance to ensure that existing U.N. entities and processes are strengthened, rather than creating duplicative functions.” Lu, also a representative for U.N. Management and Reform, said “there is already significant work being done on digital issues in the U.N. system, including at ITU, UNESCO, and UNDP. The multilateral development banks are also making significant investments to connect the unconnected.” The deliberations at the U.N. are playing out against a backdrop of rising geopolitical tensions. The U.N. has envisioned GDC as a vehicle for ensuring that digital technology benefits the world’s poor and rich alike, guards against discrimination and violence, and accelerates the world’s progress on the SDGs — a set of internationally agreed targets aimed at ending poverty and inequality. But beneath the surface of the debate over digital governance is a fierce geopolitical battle over two vastly different visions: The so-called multistakeholder model, favored by the U.S. and other Western powers, which advocates a range of players, including government officials, academics, technical experts, human rights experts, and tech companies working together to shape the rules that govern the digital world; and the other, favored by China and other authoritarian-minded governments, which prefers a digital future that is marked by increasing government control over the digital world, including AI. It has also fueled suspicion among civil society organizations that China, Russia, and other authoritarian powers have sought to exploit the U.N. diplomatic process in New York, where they wield considerable influence, to promote greater state control in the digital realm, cutting out a loose coalition of academics, internet pioneers, and tech entrepreneurs who have helped shape the norms around digital issues. “Authoritarian efforts to undo the multistakeholder model of an open internet are finding new life through the U.N.’s Global Digital Compact,” according to an informal paper prepared by a coalition of civil society groups. “Internet and democracy experts have expressed concerns that the process as currently scoped risks undermining the multistakeholder system of internet governance, and is being leveraged by authoritarian states like China to advance their vision of a state-controlled digital world.” China has sought to position itself as an advocate for low- and middle-income countries, highlighting the need to ramp up funding in digital infrastructure, involving such states in discussions on AI, and promoting technology sharing and “full and equal access of developing countries to AI technologies.” “We believe that the United Nations should be the main channel for AI governance,” a Chinese diplomat said at the opening of the digital compact talks in early April. “The process must be inclusive, transparent and respecting national sovereignty. It should be member state-driven and based on consensus.” The U.S., meanwhile, and its allies have pushed back, voicing deep skepticism over the need for the U.N. to establish new entities, contending that effective digital governance institutions already exist and should be supported, and not supplanted. The AI drive is part of a broader effort to find solutions to a crisis in trust and cooperation in the multilateral system, driven in part by the growing equality gap that followed the pandemic, and the wars in Gaza and Ukraine. “The U.N. effort to rethink the multilateral system is worthwhile, particularly given the current geopolitical tensions” over Gaza and Ukraine, Konstantinos Komaitis, a resident senior fellow at the Atlantic Council, told Devex by telephone. There is a “certain frustration in the U.N. system both from the member states, as well as from the secretary-general, that something is really not working. We need to do something.” The draft of the digital compact has some positive provisions, including frequent references to the importance of upholding human rights standards “in a way we haven’t necessarily seen beforehand,” Komaitis told Devex. “It’s a big deal.” But there are also red flags buried in the text, including the use of undefined phrases often deployed by authoritarian governments. For instance, the draft includes frequent references to promoting a “free, universal and secure” internet, a phrase frequently used by China. That is a shift from the more conventional phrase “open, global and interoperable.” “They are much more subjective. The community has not spent time defining them.” “We’re seeing a certain degree of concentration, or an attempt at concentration, of some of those governance functions within, not even just the U.N., but within the secretary-general’s office,” he said. “One of the things we have learned about the internet is that it better corresponds to decentralized forms of governance.”