AINEWS
A decade ago, he warned that superintelligence was 'probably the greatest threat to the continued existence of humanity.' Now he's briefing the UN's most powerful body.
OpenAI CEO Sam Altman will brief the 15-member UN Security Council on AI risks next week. The Council, not the General Assembly, is the only UN body empowered to impose binding resolutions, and the choice of forum carries more information than the prepared remarks.
The session is scheduled for Wednesday, September 23, during the high-level week of the UN General Assembly in New York. France, which holds the September rotating presidency, placed AI on the Council's program of work on September 14. The session is described as "open," a format the Council typically reserves for high-stakes briefings rather than working sessions.
The briefing will be the Council's second formal meeting on AI risks. The first, in July 2023, produced two positions that still frame the diplomatic landscape: China's warning that the technology should not become a "runaway horse," and US concerns about AI used to censor or repress, according to UN-watcher reporting.
The briefing is worth attention for the speaker, not the speech.
A decade before OpenAI existed, Altman wrote a short essay titled "Machine Intelligence, part 1" arguing that "the development of superhuman machine intelligence is probably the greatest threat to the continued existence of humanity." That warning predates his role at OpenAI, the launch of ChatGPT, and the company's current market position. The line is widely cited; it is also the single sentence that most clearly anchors what Altman has said, on the record, he believes about his own field.
His expected remarks on Wednesday will focus on voluntary safety work and the need for international coordination. That is the position OpenAI has carried into Senate hearings, the AI Action Summit in Paris, and now the Council chamber. It is also the position most adjacent to a frontier lab's commercial interest in keeping development moving under rules it can help shape.
A General Assembly speech is a platform. A Security Council briefing is a policy move.
The Council can adopt resolutions under Chapter VII of the UN Charter, including sanctions and, in principle, mandates that bind member states. It cannot, however, legislate AI. Its remit is international peace and security, and "existential AI risk" is an unsettled fit for that mandate. France's program of work frames the session around "existential threats from uncontrolled AI, global peace, and the welfare of children." That phrasing bridges the AI safety community's risk vocabulary and the Council's traditional portfolio.
Wednesday's session lands weeks after several AI industry leaders publicly called for a coordinated slowdown in the development of more powerful systems. Reuters did not directly name the Amodei-led call in its reporting on the briefing; the two events are adjacent, not the same, and should not be conflated.
Three signals will tell readers whether the Council treats this as a one-off or as the seed of a working agenda.
First, which Council members engage substantively with Altman in the question period, and which treat the session as theater. The US, UK, and France are likely to use the session to reinforce voluntary-coordination language. China and Russia will signal whether they accept AI as a security-domain issue on par with arms control, or push back on the framing.
Second, whether the session produces any procedural step beyond the briefing itself: a follow-up session, a presidential statement, or a request for a Secretary-General report. A presidential statement is the lightest form of Council output and is the realistic ceiling for September 23.
Third, whether Altman makes any concrete commitment beyond talking points. He may announce a new safety commitment, a new external review, or a pledge of technical access for the Council's Group of Governmental Experts. He may also stick to the prepared script. The difference will tell readers how seriously to take the moment.
A frontier-lab CEO asking the UN's binding-rules body to coordinate on AI is, on its face, exactly the development the AI safety community has been asking for. It is also the development the same community has warned against: a small number of companies writing the international rulebook for a technology they are simultaneously racing to build. Wednesday's session will not resolve that tension. It will, at most, put it on the record.