Sam Altman Addresses UN Security Council on AI Future

OpenAI CEO Sam Altman spoke to the United Nations Security Council on Tuesday, making the case for international coordination on advanced artificial intelligence. He is the first leader of a major AI lab to address this specific body.

His central argument was that AI safety is a global problem requiring a global solution. Altman urged the council to form a global body, similar to the International Atomic Energy Agency (IAEA), to oversee the most powerful AI systems. This would involve international auditing, safety standards, and a mechanism to prevent proliferation of dangerous capabilities.

The choice of venue is the news. The UN Security Council deals with international peace and security. Its permanent members include the United States, China, Russia, the United Kingdom, and France. Altman placing AI governance in this context signals a shift from a corporate responsibility issue to a matter of state security.

Altman outlined specific risks. He pointed to AI’s potential to help create novel biological weapons. He also warned about AI-enabled cyberattacks that could disrupt critical infrastructure. The implication is clear: these are threats that cross borders and could destabilize nations, hence the Security Council’s attention.

From a technical perspective, Altman highlighted two key areas. The first is alignment, the process of ensuring an AI system’s goals match human values. The second is governance, the external oversight that constrains a system’s actions. He argued that alignment alone is insufficient. Even a perfectly aligned system could be misused or deployed in unstable geopolitical environments.

This aligns with ongoing research in AI safety. Labs like OpenAI and Anthropic are developing methods to interpret and control advanced models. But Altman is now arguing that technical safeguards must be paired with political ones. He suggested a licensing regime for the development of models above a certain compute threshold.

Reactions from the AI community are divided. Some researchers and policy experts welcome the call for a global body. They argue that a race to the bottom between nations is likely without guardrails. Others are skeptical. They point out that a body like the IAEA works because nuclear material is physical and trackable. AI models are code. They can be copied and run secretly on a gaming PC.

Critics also note a tension in Altman’s position. He leads a company racing to build the most capable AI systems in the world. Advocating for international controls could be seen as an attempt to lock in advantage for established labs that can afford compliance. Smaller developers and open-source communities might be squeezed out by strict global licensing.

The discussion also touches on the compute supply chain. Advanced AI chips are manufactured in a handful of countries, primarily by TSMC in Taiwan. This existing choke point could make some forms of hardware-based governance feasible. But it also creates a point of tension between major powers.

Altman’s proposal puts specific pressure on nations like China. They are investing heavily in domestic AI development and have their own visions for governance. Getting agreement from all Security Council members on an international AI agency would require navigating deep geopolitical divides.

The path forward is unclear. The UN Security Council has no immediate mechanism to create a tech regulator. It first needs to agree on a resolution to study the issue further. Altman’s appearance serves to elevate the timeline and inject corporate perspective into a slow-moving diplomatic process.

Can a regulatory body for AI be established before the technology itself outpaces all oversight? That is the concrete challenge laid before the council.

Source: Sam Altman’s remarks at the United Nations Security Council