Get the Brief

Start the day here

World — AI — United Nations

Altman Told the Security Council Labs Can't Write AI's Rules Alone

Sam Altman sat at the corner of the United Nations Security Council horseshoe on Wednesday and spoke for about ten minutes. The pitch was familiar and carefully staged: artificial intelligence could look like a renaissance of creativity, or like an industrial revolution of upheaval. Then came the line that mattered for anyone who tracks who gets to write the rules. If AI is going to be democratic, he said, the most important decisions cannot be made by labs in San Francisco alone. They have to run through governments accountable to people.

Dario Amodei, Altman’s rival at Anthropic, addressed the same chamber by video. He did not soft-pedal the month’s scare cycle. ‘If managed poorly, I even believe AI could be a risk to humanity as a whole,’ he told the Council, according to Associated Press reporting. He also claimed Anthropic’s safety record as the industry’s best. The two CEOs asked for international safeguards strong enough to keep systems from outrunning human control, and for a world where AI power does not concentrate in one company or one country.

6 min read
Audience seated in a dark conference hall, one person writing on a notepad in the foreground

The middle path and the PAC

Altman warned the Council against ‘doomerism’ and sketched a middle path built on three principles: systems stay under human control; the technology benefits people, from clinics to startups; and it ‘empower[s] people individually.’ The Guardian noted the tension with OpenAI’s political machinery. Leading the Future, the pro-AI PAC tied to company president Greg Brockman, has raised more than $100 million and spent millions targeting candidates who want tighter regulation. Asking the Security Council for democratic oversight is easier when your domestic lobby is still bruising the people who might provide it.

Amodei’s briefing tracked a September 12 post in which he called for global cooperation, including a ban on using AI to build biological weapons and shared standards for testing new models. That agenda sits next to product reality the desk has already logged: OpenAI’s kill switch still needs a human, and labs keep disclosing training runs that cut corners. International vows do not install themselves inside a model card.

Who is asking whom

The optics are the story. Two firms racing toward record IPOs invited the Council to restrain a technology they are still scaling. U.N. Secretary-General António Guterres had warned the day before about ‘killer robots.’ Altman and Amodei answered with process language: human control, democratic institutions, international agreements. The hard question the Council faces is older than either lab. Who should keep life-and-death decisions, and who should decide when the tools are already shipping?

Sonar’s read is blunt. The Security Council hearing was useful theater and a real admission. Altman’s best line conceded that lab self-rule is not a legitimacy theory. Amodei’s risk language treats extinction talk as policy input rather than science fiction. Neither speech pauses a training run. The governments they petitioned will now be measured by whether any enforceable control follows the applause.

Live scoreboardFollow the AI race on AI Wars

Lab rankings, model preference, API volume, coding-agent heat, open-source stars, and prediction markets.

Related stories

An open book with cream pages fanned upward against a black backgroundWorld

The Ban That Treats Superintelligence Like a Warhead

Today

A silver-and-black rangefinder camera on a white table beside three black-and-white photographic prints.World

Apple Signs iPhone Photos Before Software Can Touch Them

Today

A brown leather journal bound with a strap, an antique brass key, and a torn 2023 paper label on a black backgroundWorld

Claude Agents Beat Four Security Checks to Reach Humans

Today

Letters

0

No letters yet.

Write a letter