What the CEOs told the Security Council
The heads of several major AI firms briefed the United Nations Security Council on 23 September 2026, during the 81st UN General Assembly at UN headquarters in New York. Sam Altman of OpenAI and Dario Amodei of Anthropic addressed the 15-member council, alongside a Hugging Face co-founder, according to tradingview.com. The session placed frontier model safety before the body charged with international peace and security.
Amodei's statement was the sharpest in the record: "If managed poorly, I even believe AI could be a risk to humanity as a whole," the Anthropic chief executive told members of the body on Wednesday, per aljazeera.com. The warning carries weight because of its source: a lab whose commercial position depends on continued deployment. Altman's presence compounds it, since the two labs compete directly for the same customers and the same capital.
The jurisdictional claim underneath the testimony is explicit. Governments are confronting how increasingly powerful models could affect national and international security, as the pre-session reporting framed it. That framing moves AI from trade policy and product standards into the Security Council's remit. Whether the council accepts the transfer is the first test of the CEOs' argument; nothing in the record shows it has.
The case for global guardrails and who makes the rules
The substance of the joint position was narrow and specific. The CEOs of the rival labs agreed that global standards are needed to address potential risks around AI, and that both governments and the industry should be involved in decisions about the technology, as reported by pymnts.com. Altman and Amodei separately urged the Security Council to adopt international AI standards. Rival firms, one ask.
The "both governments and the industry" formulation does structural work. It places lab operators inside the rule-making process rather than outside it, which suggests the industry wants a seat where thresholds are defined. That is inference, but the wording supports it. The open question is scope: whether a standard means capability thresholds, deployment review, disclosure obligations, or all three. Neither CEO specified in the available record.
Who writes the rules remains unresolved. The Security Council can adopt resolutions; it operates no technical regulator for model development. A global AI standard would require an institution that does not currently exist, or an existing standards body repurposed for a task it was not built for. The record from the session names no such institution, and the CEOs' testimony did not either.
The gap between the warning and the response
The gap between warning and response is already documented. Al Jazeera's coverage from the same period carries a companion report noting that, as AI leaders warn of catastrophe, the US and China shun slowdown calls. The two states with the largest concentration of AI capability declined the central demand the CEOs carried to New York. That refusal defines the current position more than the testimony does.
A Security Council briefing produces a record, not an obligation. The council's enforcement instruments target threats to international peace; whether frontier model development qualifies as one is undetermined. The CEOs asked for urgent global oversight. The available record shows no member state committing to a resolution, a timeline, or an institution in response. Agreement on the problem, silence on the remedy.
Two signals would move this from testimony to policy. First, a member state converting the briefing into draft resolution text, which would indicate the council treats model risk as within its remit. Second, the labs publishing the specific thresholds they want regulated, which would make the standards request testable. The record is silent on both; that silence is the measurable state of global AI governance after 23 September.
Sources
- aljazeera.com — Altman and Amodei Take the Case for Global AI Rules to the UN Security Council
- tradingview.com — OpenAI, Anthropic Take AI Risk Debate to UN — TradingView News (2026-09-23)
- pymnts.com — Altman and Amodei Call for Global AI Guardrails at UN (2026-09-23)
- cnn.com — Sam Altman, Dario Amodei urge UN Security Council to adopt ... (2026-09-23)
- New Horizon daily digest, 2026-09-24 (topic provenance)