Altman and Amodei brief the UN Security Council on AI safety

OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei addressed the United Nations Security Council on September 23, delivering separate briefings on AI safety during the body's high-level week at UN headquarters in New York.

The two men lead the two most prominent US-based AI labs working on frontier models. Their joint appearance before the 15-member council — an institution that traditionally convenes around armed conflict, sanctions, and counterterrorism — signals that advanced AI development has reached the diplomatic agenda of the UN's most security-focused organ.

"We have a choice in front of us," Altman told the council. "AI can either be more like a new renaissance of creativity and discovery, or more like a new industrial revolution of upheaval and disarray."

The Guardian first reported the briefings. The session coincided with the United Nations General Assembly high-level week, which drew heads of state, diplomats, and technology executives from around the world.

Why the Security Council, not a tech forum?

The choice of venue carries the story. AI governance typically plays out at the OECD's AI Policy Observatory, the G7's Hiroshima AI Process, or the EU's AI Office. Those bodies focus on standards, transparency, and rights.

The UN Security Council focuses on threats to international peace. Bringing frontier AI executives before it implies that member states now treat advanced AI as a potential security matter on par with weapons proliferation or cyber operations.

The council's tool kit is narrow. Its members can issue presidential statements, request secretary-general reports, and pass resolutions. They cannot regulate companies directly. Their leverage runs through the diplomatic weight they lend to subsequent UN processes.

A briefing on AI safety, even without a binding resolution, can push the UN secretary-general to commission a new study on frontier AI risks — modeled on existing reports to the council on autonomous weapons and cyber operations.

Why two separate briefings?

Altman and Amodei appeared in distinct sessions rather than a joint appearance. That format suggests the council wanted each company's perspective individually rather than a unified industry pitch.

The two men run rival organizations with different positions on deployment speed, government engagement, and safety review. OpenAI, the San Francisco company behind ChatGPT, has emphasized broad access and rapid iteration. Anthropic, founded in 2021 by former OpenAI researchers including Amodei, has built its public identity around model safety research and its Claude family of assistants.

The separate slots gave each CEO room to address the council without negotiating a shared industry line. That matters. OpenAI and Anthropic have publicly disagreed on pre-deployment testing requirements, the pace of frontier model releases, and the appropriate level of regulatory oversight. The council could hear those competing views side by side.

What did Altman tell the council?

Altman's brief quote framed the AI safety debate as a fork between two outcomes. The "renaissance" frame evokes open creativity and scientific acceleration. The "industrial revolution" frame flags mass labor displacement, geopolitical competition, and concentration of power.

The phrasing tracks the binary Altman has used at prior public appearances, where he argues that AI policy will shape whether the technology produces widespread prosperity or widespread harm.

His audience this time was different. The Security Council's 15 members — five permanent and ten rotating — include China, Russia, the United States, France, and the United Kingdom. Speaking directly to those governments gives Altman and Amodei access to regulators who often hear about AI through intermediaries and second-hand briefings.

What did Amodei bring to the briefing?

The Guardian's initial report did not detail Amodei's remarks. Anthropic has published research on catastrophic risk, model interpretability, and responsible scaling — work that shapes how the company frames AI dangers in public forums.

His appearance offered the council a contrasting voice. Where OpenAI emphasizes product shipping, Anthropic has publicly pushed for layered safety reviews before each new model generation. Amodei has argued in past writings that frontier models could enable catastrophic harm if deployed without adequate safeguards.

The two briefings, taken together, gave the council a window into the divergence between the two US labs that define the frontier AI sector today.

What does the industry stand to gain?

For OpenAI and Anthropic, the appearance carries strategic weight. The companies that brief the council today shape the rules their products will live under tomorrow.

Speaking at the UN also locks both firms into a continuing diplomatic relationship. Follow-up briefings become routine once CEOs testify. The labs now have a standing channel to the people who decide on export controls, compute rules, and international treaty language.

That dynamic benefits established labs with policy staff and safety teams. Smaller open-source projects and non-US AI companies cannot easily match the institutional access. The precedent could harden into a closed-door governance process defined by the largest US labs.

What is the council's track record on tech?

The Security Council has engaged with technology before. Its past sessions on cyber operations helped frame the UN's work on state-sponsored hacking. Council attention to autonomous weapons has pushed the Convention on Certain Conventional Weapons to revisit lethal autonomous systems.

AI is now on a similar trajectory. The September 23 briefings placed frontier models on the council's radar in the same way offensive cyber tools and autonomous arms reached the body years ago.

Member states will read the briefings as a signal to invest in AI policy staff, treaty language, and intelligence assessments. The corporate briefings give diplomats specific framings they can adopt or contest in subsequent negotiations.

What happens after the briefings?

The September 23 sessions do not yet amount to regulation. They create a precedent: the heads of frontier AI companies have now personally addressed the body charged with global security on AI safety.

The next markers to watch are a presidential statement from the council, a secretary-general report on frontier AI risks, and any return of the topic during the General Assembly's main session later this year. Each step would tighten the diplomatic scope around the industry.

For now, the briefings stand as a signal that the era of voluntary AI governance is closing — and that the world's most powerful security body has decided to pay attention.