Bill Gates says AI needs federal law — not self-regulation — to prevent mass casualties
Bill Gates has called for mandatory federal oversight of artificial intelligence in the United States, arguing that leaving safety rules to the tech companies themselves is not enough. Speaking to NBC News, Gates said the US "absolutely" needs legislation requiring independent safety standards and algorithmic monitoring. Meanwhile the EU began enforcing its own AI rules in August — and American labs are already subject to those fines.
The claim
Gates did not reach for science-fiction scenarios. He focused on present-day risks: bad actors using existing systems to cause harm at scale. "AI is certainly powerful enough to drive events that cause a billion deaths," he told Meet the Press. He wants regulators — not developers — to set the floor, drawing a parallel to aviation: you don't let the aircraft manufacturer decide whether its own plane is safe to fly. Microsoft President Brad Smith has used the same analogy publicly.
The position puts Gates in line with Anthropic's Dario Amodei, who in September proposed a voluntary slowdown on frontier-model development, and OpenAI's Sam Altman, who supports international safety standards. Mark Zuckerberg sits on the other side: Meta's position is that each lab should assess its own risks internally, with no industry-wide coordination required.
That Anthropic stance is worth watching closely — the company recently Anthropic built a secret biology lab — and relaxed its own bioweapon safeguards after opening a secret biology lab, which critics say illustrates exactly why self-regulation falls short.
States vs. Congress
With no federal bill in sight, individual states are moving first. California Governor Gavin Newsom signed an executive order on September 18 requiring an expert panel to design an emergency "kill switch" for the most powerful AI models by November 16. Maryland and New York are drafting their own measures.
The EU comparison
The gap between US debate and European action is stark. The EU AI Act's enforcement powers went live on August 2, 2026, allowing the European Commission to fine AI providers up to 3% of global annual revenue for safety violations. US labs operating in Europe are already facing that compliance cost — without a matching domestic requirement. Gates' proposed framework would essentially replicate what Brussels is already doing.
Gates was careful to reassure industry: sensible rules should not freeze development. The "Wild West" period, he argued, needs to end before something irreversible happens — not after.
