Bill Gates says AI is powerful enough to help trigger a billion deaths, and he does not trust companies to police themselves.
Gates argued that safeguards and monitoring cannot be optional extras left to individual AI labs to decide on their own. He pointed to the European Union's AI Act as the model other governments should follow: the bloc has required risk safeguards for AI systems since August 2025. Since August 2, 2026, the European Commission has had the power to fine AI model providers that violate those rules. Gates framed that enforcement authority, not the safeguards themselves, as the missing piece in most countries' approach to AI oversight.
Gates is one of the highest-profile tech figures to say plainly that voluntary AI safety pledges are not working. His argument leans on the EU's willingness to attach real financial penalties to non-compliance, something the US and most other governments have avoided so far.
Every major AI lab has made safety promises in recent years; whether any of them change behavior without a regulator that can actually fine them is the question Gates is really asking.