Bill Gates says unregulated AI is powerful enough to ‘cause a billion deaths’ -- and wants to make that case to Trump in person

Bill Gates spent Sunday's episode of NBC's Meet the Press making the case that 2026's loudest AI warnings have not been loud enough. Asked by host Kristen Welker about the risks of unregulated AI development, the Microsoft co-founder said the technology is powerful enough to drive events that... cause a billion deaths, adding that "there's never been a weapon as powerful as the combination of people with ill intent using the latest AI tools." It is the starkest public warning yet from a technologist of Gates's stature, delivered two weeks after President Trump dismissed similar alarm from AI-lab chief executives as "a hoax."

“AI is certainly powerful enough to drive events that... cause a billion deaths.” — Bill Gates, on NBC's Meet the Press, Sept. 27, 2026

Gates was direct about what he thinks is missing: government involvement, not industry good faith. "No one thinks self-regulation is enough," he said, arguing "you need law enforcement and the politicians to get into the discussion about what safeguards and monitoring look like." He also rejected the idea that an emergency shutdown mechanism is sufficient on its own -- "it's not enough to have a kill switch," he said, arguing that regulators need "insight and records of what's being done," because a kill switch alone "would not prevent these tragedies."

Sunday's remarks were not a one-off. Gates has been escalating steadily through 2026: in January he warned that a non-government group could use open-weight AI tools to design a bioterrorism weapon; in August he argued in a lengthy essay that society faces "a very high chance of a net negative outcome" on AI's current path, and that managing its downside should be "the world's top priority"; Sunday's 'billion deaths' framing is the sharpest point yet on that same line. (The trajectory runs opposite to the industry's usual pattern -- founders and early backers tend to grow more bullish as money floods into their field, not less. Gates's own foundation pledged $1 billion toward beneficial AI applications the same year, so this isn't a retreat from the technology so much as a widening gap between how much upside and how much downside he now sees in the same thing.)

That puts Gates in company he didn't create. Anthropic's Dario Amodei and OpenAI's Sam Altman spent much of September making a related case -- Amodei has called for the industry to 'pace' its own development, and both he and Altman backed a public call for stronger safeguards that Trump and adviser David Sacks rejected on the record two weeks earlier, dismissing the warnings as a "SICK conspiracy." Gates's language is more alarming than either lab chief's own public framing, but his underlying ask -- federal involvement, not voluntary restraint -- is the same one Trump has already said he doesn't accept.

None of the three positions above maps neatly onto industry self-interest. Gates has no frontier lab to protect and no product to sell into a slower-moving market -- if anything, a heavily regulated AI industry could slow the return on Microsoft's own multibillion-dollar OpenAI stake. (A lab CEO calling for caution can be read as tactical -- credentialing the industry as too dangerous to disrupt without its own input, a criticism Sacks made explicitly of Amodei and Altman. A co-founder with no active AI company to protect doesn't carry the same read.) That is part of why Gates's intervention lands differently. The administration had not, as of Sunday, issued a specific response to his remarks; its existing position, delivered two weeks earlier, is that current "CRIMINAL and REGULATORY power" is already sufficient and that further safeguards would slow an AI race the US needs to win outright.

Two government-level developments the same week complicate the domestic picture further. The White House said Sept. 26 that Trump and Xi Jinping had agreed to open a bilateral channel for reporting AI-related incidents between Washington and Beijing -- a government-to-government safeguard that is, notably, a form of exactly the "insight and records" monitoring Gates says domestic AI policy still lacks. New York City's Council had introduced its own kill-switch and whistleblower-bounty legislation the day before that, summoning OpenAI and Anthropic to testify at a hearing on the bill. Gates's warning, delivered the same week, gives both efforts a validator with no seat at either table -- he is not a lawmaker, a company executive, or a foreign official, which is precisely what makes Trump's "hoax" framing harder to apply to him specifically.

Gates's own comparison may be the most useful frame for judging what happens next: he called reaching international agreement on AI "more difficult" than Cold War-era nuclear arms negotiations -- not impossible, but slower and messier than either side's public rhetoric suggests. The concrete tests are the same as with any other pledge in this fight: whether the meeting with Trump he says he wants actually happens, and whether any of it produces legislative text rather than another interview.