The UN enters the chat: dueling resolutions and a compact

Lesson 1 of 4 in The United Nations and the Treaty Track: Global Digital Compact and the CoE Convention.

For years the UN watched AI governance happen elsewhere — in Paris at the OECD, in Brussels, in G7 communiqués. Then, in a single eighteen-month stretch, the General Assembly produced two consensus resolutions, a flagship advisory report, and a compact committing all member states to build new AI institutions. Understanding why the UN track matters requires one uncomfortable premise: the OECD represents 38 rich countries, the G7 seven. Most of humanity was governed by AI rules it had no seat in writing. The UN is the only room where that changes.

The opening move came on 21 March 2024: resolution A/RES/78/265, the first General Assembly resolution on AI — US-led, with more than 120 co-sponsors, adopted by consensus. Its title tells you the framing: safe, secure and trustworthy AI for sustainable development. Four months later, in July 2024, came the counter-move: a China-led resolution on AI capacity-building, also adopted by consensus — with the US among the co-sponsors. Two superpowers, two resolutions, each careful to join the other’s. That is what AI diplomacy looks like: competition conducted entirely in the language of cooperation.

US-led: 78/265 (Mar 2024)

Frame: safety, rights, and trust. The resolution urges member states to promote safe, secure and trustworthy AI systems, respects human rights language throughout, and ties AI to the Sustainable Development Goals. The diplomatic achievement was consensus: China and Russia joined rather than forcing a vote — because the text asks nothing binding of anyone.

Read it as the OECD/G7 vocabulary arriving at UN scale: the same trustworthy-AI framing you met in the soft-law module, now endorsed by every member state.

China-led: capacity-building (Jul 2024)

Frame: development and the digital divide. China’s resolution emphasizes helping developing countries build AI capacity — infrastructure, skills, data — and calls for international cooperation so the Global South is not locked out of the AI economy.

Read it as positioning: China presenting itself as champion of the majority world, implicitly contrasting with a Western governance conversation about frontier-model risk. The US co-sponsored — declining to concede the Global South audience. Both resolutions passed by consensus; neither binds; both were written for the audience of the other 150 states.

While the resolutions traded frames, the Secretary-General’s High-Level Advisory Body on AI — 39 experts from government, industry, and academia — was drafting the intellectual blueprint. Its final report, Governing AI for Humanity (September 2024), made seven recommendations, anchored by two institutional proposals: an independent international scientific panel on AI (modelled loosely on the IPCC) and a regular intergovernmental policy dialogue on AI governance. The other five — a standards exchange, a capacity-development network, a global AI fund, a data-governance framework, and a small AI office in the Secretariat — met the usual fate of UN report recommendations: partial, slower uptake.

The vehicle for the two big ones arrived within weeks. At the Summit of the Future (September 2024), member states adopted the Pact for the Future, with the Global Digital Compact (GDC) annexed to it. The GDC’s Objective 5 commits states to enhancing international governance of artificial intelligence for the benefit of humanity — and specifically to establishing the Scientific Panel and a Global Dialogue. A last-minute Russian amendment to defer the package failed decisively, and the Pact was adopted without a vote. Remember the GDC’s legal character, because it recurs in the final lesson: a compact is a political commitment, not a treaty — nothing in it binds anyone, yet it is the reason two real institutions now exist.

Key terms: unga resolution, global digital compact, pact for the future, consensus adoption

Interactive checkpoint quiz (2 questions) — open this page in a browser to take it.