Speaking to reporters in New York on 16 September 2026, ahead of the 81st session of the General Assembly, United Nations Secretary-General António Guterres called for two new instruments to sit alongside the machinery the organisation has already built for artificial intelligence. The first is an AI Child Safety Pledge. The second is a Global Fund for AI that the UN says would help developing countries build AI infrastructure and capacity, according to UN News and to the organisation’s own digital and emerging technologies office, which published the appeal under the heading of capacity-building gaps.
The existing machinery is worth recalling, because it sets a realistic expectation for how quickly anything new would work. UN News records that the Global Digital Compact, adopted by member states in 2024 as part of the Pact for the Future, asked for two mechanisms: an independent scientific panel on AI and an annual global dialogue on AI governance. The General Assembly formally established both in August 2025. The Independent International Scientific Panel on AI released its first preliminary report on 1 July 2026, concluding that capabilities are advancing quickly in reasoning, coding and science while also identifying risks that include misinformation, discrimination, privacy violations, cyberattacks and a possible “superintelligence” that could become difficult to control. Two years therefore passed between the call for the mechanisms and the first substantive evidence product they produced.
Guterres framed the new proposals inside a broader argument about coordination. He told the press conference that “national action is essential, but global coordination is indispensable”, and he argued that voluntary efforts to slow AI development will not be enough if they are isolated, unverifiable or unevenly applied, as reported by UN News. He compared frontier AI competition to the Cold War arms race and said states with the greatest frontier capacity, including the United States and China, should establish mechanisms of contact and exchange information to avoid what he called a race to the bottom. Anadolu Agency quoted his call for guardrails that make AI “safe, transparent, accountable, with human dignity at the center”. Xinhua reported the same remarks with the emphasis on his claim that the world faces three existential threats, which he listed as runaway artificial intelligence, the climate crisis and deepening inequalities.
What the two proposals do not yet specify
The announcement as published leaves the practical architecture undefined. The UN’s own account describes the fund’s purpose without naming a target size, a contributor list, a governing board or the agency that would assess applications and release money. A pledge, by definition, is voluntary, which sits awkwardly beside the Secretary-General’s own statement in the same press conference that voluntary commitments fail when they cannot be verified. If unverifiable voluntarism is the problem he identified, then a corporate pledge on child safety answers it only if the pledge carries an audit mechanism, and none was described.
The same gap applies to the fund’s incentives. UN News states that the UN’s role will be to bring governments together, provide a scientific basis for discussions and identify areas for international cooperation. That is a description of convening, and convening does not usually require a treasury. Once an institution controls disbursements for AI infrastructure in developing countries, it acquires leverage over which technical standards, vendors and governance models those countries adopt, and the recipients have limited means to contest the conditions. Those are ordinary questions about money and power, and the announcement did not answer them.
Where the political conflict actually sits
The proposals also landed in an open disagreement between the UN and Washington. Channel News Asia, carrying Reuters copy, reported that Guterres warned of risks that cannot be ignored and that this put him at odds with President Donald Trump, who has argued that existing safeguards are sufficient. The Independent, carrying Associated Press copy, reported his appeal to global AI competitors on the grounds that the world cannot afford a race to the bottom on AI safety. Both accounts point to the same practical obstacle: a fund and a pledge require the participation of the country that hosts most frontier developers, and that country’s administration has publicly taken the opposite position on the need for new rules.
How the outlets framed it
UN News presented the fund and the pledge as capacity-building for developing countries and cast the organisation’s own role as convening governments and supplying a scientific basis for their discussions, which keeps the story in the register of technical housekeeping. Channel News Asia, running Reuters copy, framed the identical remarks as a confrontation, reporting that the Secretary-General was at odds with President Donald Trump over whether existing safeguards suffice. Xinhua moved the emphasis again, foregrounding the three existential threats and the duty of governments to act. The difference matters because the first framing makes the fund sound administrative, while the second and third reveal it as a contested political instrument whose success depends on states that disagree about whether the problem exists.
The measurable tests over the coming high-level week are straightforward. Observers can watch for a named set of contributing states, a disclosed governance structure for disbursement, and any verification language attached to the child safety pledge. If the General Assembly week ends with the proposals still described only in terms of ambition, the record from 2024 to 2026 suggests the first concrete output would arrive somewhere near the end of the decade.
