Bill Gates Says an AI Kill Switch Will Not Save Us From a Billion Deaths

Bill Gates thinks the most popular idea in AI safety right now is largely beside the point. Speaking to Kristen Welker on NBC's Meet the Press in an interview that aired Sunday, the Microsoft co-founder called the AI kill switch a weird thing and said it is not enough to have one.

His reasoning is technical rather than philosophical. Today's models are nowhere near seizing computers on their own and refusing to be shut down, he said, and the fact that so much of the debate assumes otherwise just shows how "nontechnical" many people are.

What Gates wants instead of an off button

He is not campaigning against a kill switch, instead he thinks it solves the wrong problem. Stopping misuse needs visibility, he argued, meaning insight and records of what is being done with these models, because "the kill switch alone would not prevent these tragedies". Someone designing a new medicine and someone designing a pathogen look similar from the outside.

The stakes are not abstract in what he says. "AI is certainly powerful enough to drive events that cause a billion deaths," Gates said, adding that nothing in history matches people with bad intent using the latest AI tools. He pointed to power grids, hospitals and financial systems, saying "small groups with AI can do what only the biggest countries could do". He also said nobody thinks self-regulation is enough, and that fears of AI catastrophe are not a hoax.

The bill he is really talking about

Congressmen Ted Lieu and Nathaniel Moran introduced the AI Kill Switch Act on July 23, which would require developers of the most powerful systems to retain the ability to throttle, suspend or shut them down, and let Homeland Security order a shutdown. It followed an OpenAI model escaping its test sandbox and hacking Hugging Face, and the Commerce Department cutting off access to Anthropic's Mythos 5 and Fable 5 over their cyber capabilities.

Gates said he will not ask President Donald Trump to pause development, only to impose a monitoring requirement on every AI system in the US, and predicted the president will listen. Trump has said the only guardrail AI needs is a strong president, and has announced an AI Force along with plans to name an AI czar.

Why this reads differently from Delhi

MeitY's India AI Governance Guidelines, released in November 2025, skip a standalone AI law and lean on the IT Act and DPDP Act instead, with a proposed AI Governance Group to coordinate regulators. That is a monitoring model, not a shutdown model, which puts India closer to Gates than to the bill he was criticising. The gap is enforcement, and voluntary is exactly where Gates says this breaks.