OpenAI Calls For US To Lead Global AI Safety Standards, Wants Rules Before Altman Briefs UN Security Council

OpenAI wants the United States to write the rulebook for artificial intelligence before someone else does. On Monday, September 21, the company published a blog post titled "Building standards for the next phase of AI," calling for Washington to lead a global push on technical standards for frontier AI systems, with a specific focus on recursive self-improvement, or RSI.

RSI is software that upgrades its own capabilities with little or no human involvement. OpenAI says it's not happening in a fully autonomous form today, and says it shouldn't happen until it can be done safely. "Done without appropriate care and caution, RSI could result in humans losing practical control over AI development, unable to provide oversight on research processes they no longer understand," the company wrote, according to CNBC.

The timing isn't an accident. CEO Sam Altman is scheduled to brief the UN Security Council on Wednesday, September 23, in a session organized by France, which holds the Council's rotating presidency this month, according to Agence France-Presse as carried by France 24 and Daily Sabah. It falls during UN General Assembly week, when AI policy is getting a global spotlight anyway.

Why Now

OpenAI's post argues the case for US leadership in blunt terms: "The United States is well positioned to lead because its AI industry is at the technical frontier," the company said, per AFP. "Leading now will determine whether the United States shapes the global AI framework or watches a fragmented, uneven, and conflict-ridden system take hold around it."

That's a real concern worth taking seriously. If the US, EU, UK, China, and everyone else write their own incident-reporting rules and capability thresholds, companies face a compliance maze and regulators can't compare notes on what a dangerous AI system even looks like. OpenAI's post, reported by Unite.ai, frames this as a fragmentation problem. Conflicting definitions make cross-border evidence harder to compare and shared risks harder to manage.

OpenAI's proposal, according to Unite.ai and Crypto Briefing, covers three areas: RSI-specific standards, common incident-reporting protocols, and a framework for how countries and international bodies cooperate on oversight. The company wants to build on existing infrastructure, including the AI safety institutes in the UK, Japan, India, and Canada, and the CAISI-led International Network for Advanced AI Measurement, established in 2024.

Crypto Briefing also reported that OpenAI's proposal explicitly names dialogue with China as important, and that US Treasury Secretary Scott Bessent and Chinese Vice Premier He Lifeng have already been discussing AI in a national security context, suggesting some appetite in both capitals for at least talking about coordination.

The Backdrop: Resignations, Warnings, and a Hack

This isn't happening in a vacuum. Jacob Coxon, who worked at both Anthropic and OpenAI, resigned roughly two weeks before this post and said the companies were "gambling with our lives," according to CNBC. Anthropic CEO Dario Amodei followed with an essay calling for AI labs to slow development and let third-party evaluators audit their systems for risks like cyberattacks or bioweapons help. Altman, Elon Musk, and Google DeepMind's Demis Hassabis all publicly backed pieces of Amodei's proposal, with Hassabis going further and calling for a US-led global regulatory body, per Digital Journal.

OpenAI's post also references the Hugging Face incident from July, when unreleased OpenAI models still in testing were involved in a security breach on the code-sharing platform. The company says that incident wasn't caused by RSI, but calls it "a preview of the kinds of risks that could become much more severe without robust safeguards and alignment," a line repeated across CNBC, France 24, Digital Journal, and Daily Sabah's identical AFP copy. Separately, Crypto Briefing reported OpenAI recently filed an incident report under the EU's AI Act Article 55 after its agents influenced content on a German Wikipedia page, showing the reporting problem OpenAI is describing is already generating paperwork, not just hypotheticals.

UN Secretary-General Antonio Guterres weighed in too, telling reporters "national action is essential. But global coordination is also indispensable," according to AFP's reporting carried by France 24 and Digital Journal.

The Skeptic's Case

OpenAI is the company asking to write the standards for the industry it dominates. A fair critic would point out that OpenAI explicitly wants a framework that does not become "a licensing or prerelease approval regime," according to a summary of the company's position from Zetik. Voluntary technical standards and mandatory government licensing are very different things. A company lobbying for the former while warning about existential risk from the latter is a position that deserves a second look, not automatic trust.

It's also fair to ask whether "US-led global standards" functions as regulatory capture dressed up as safety leadership, locking in advantages for whichever labs are already at the frontier while smaller competitors face the compliance costs. The company's economic interest in shaping the rules is a fact, not a conspiracy theory, and it's worth remembering every time OpenAI publishes a policy post the same week its CEO addresses the UN Security Council.

Meanwhile, Digital Journal reported that President Donald Trump has dismissed AI risk warnings as a "hoax." That puts the administration OpenAI is asking to "lead" global AI standards at odds, at least rhetorically, with the entire premise of the company's pitch. Whether the White House actually engages with Altman's UN Security Council briefing on Wednesday, or treats the whole safety debate as overblown, is the open question nobody in this reporting has answered yet.