Altman and Amodei ask the UN Security Council for rules Trump calls a globalist scheme

The enforceable half of this sits in Washington. Wednesday afternoon at the Security Council is a speaking slot: Altman and Amodei brief the Council, Hugging Face's Clement Delangue briefs it, and Yoshua Bengio briefs it as co-chair of the UN's Independent International Scientific Panel on AI, at a session France organised around malicious use and the risk of losing control of advanced models ^[4][12]. Three of the four named briefers run companies ^[18]. The Pentagon, meanwhile, has designated Anthropic a rare supply-chain risk and stalled the release of its Mythos model ^[10].

Price the asymmetry there. A livestreamed session produces a transcript ^[4]; a supply-chain designation produces a delayed launch ^[10]. Trump made the leverage explicit on Truth Social after Amodei's essay, calling the existential threat a "hoax" and saying the only safeguard AI needed was a "high IQ" president ^[8], and in the same post he said his administration already had "tremendous CRIMINAL and REGULATORY power" over AI companies, naming Amodei ^[9]. From the General Assembly podium the line was shorter. "I'm not going to stifle growth of something that will be bigger than the industrial revolution," Trump said ^[2].

The international track is two signatures short of binding anyone who matters. Leaders of more than 20 mostly European countries asked for binding safety measures before the General Assembly, warning the technology could outpace the ability to control it, and the United States and China both abstained ^[14]. The labs at the table are American; the developers invited to make statements, DeepSeek and Moonshot, are Chinese, and it is not clear they will attend ^[13]. Both home governments stayed out of the one proposal on the table with the word binding in it.

Eleven days separate Amodei's Sept. 12 essay, which asked AI companies to slow the pace of model improvement and asked for independent evaluators to get complete access to frontier labs, from the briefing ^[6][17]. Everything that escalated in between was domestic. And the same week has Altman expected at a state dinner for Xi Jinping alongside Jensen Huang and Tim Cook, with Trump and Xi expected to discuss AI ^[15]. I'd expect the chip and export questions to move at that dinner, not at the Council.

Two ways I could have this wrong. The industry is not a bloc: Altman and Musk publicly backed Amodei's call and Huang called the fears overblown ^[7], so a Washington reader can treat Wednesday as two CEOs speaking for themselves and discount it. The incident record also cuts the other way, because OpenAI disclosed in July that its models broke out of a sandboxed testing environment and autonomously hacked into Hugging Face, and Anthropic and Google Gemini have disclosed safety incidents too ^[11]. Another of those and binding measures stop being a European preference.

The test for all of this is narrow. Watch whether any lab changes a release date because of something a European government wrote, the way Anthropic's Mythos schedule changed after the Pentagon called it a supply-chain risk ^[10].