Anthropic's CEO wants the UN to draw one bright line around AI: no help building bioweapons.
Dario Amodei addressed the UN Security Council by video link today, becoming the third speaker at a meeting on AI and global security. He called artificial intelligence the most important security issue the world faces right now. Rather than pushing for sweeping AI treaties, Amodei asked the Council to start small: a specific, enforceable ban on using AI systems to help design or produce biological weapons. He framed it as a first step, not a finished framework.
Global AI governance has mostly stalled at the level of voluntary pledges and national guidelines, with no binding international rules in force. A narrowly scoped ban on AI-assisted bioweapons research is the kind of proposal that could plausibly get sign-off from rival powers who agree on little else, since it targets a shared catastrophic risk rather than competitive advantage. It also lets Anthropic put a concrete ask in front of a body that, unlike national regulators, can in theory bind every member state at once.
Getting the Security Council to agree on anything is hard enough without adding AI to the list, and Amodei's pitch says nothing about how such a ban would actually be verified or enforced.