Bill Gates says AI kill switch alone cannot prevent 1 billion deaths

Bill Gates said requiring artificial intelligence to have safeguards and outside monitoring is completely irresponsible, and warned that the technology is powerful enough to drive events causing 1 billion deaths. The Microsoft co-founder spoke in $1's Meet the Press, parts of which aired on Sunday, and argued that a kill switch alone cannot address the danger.aljazeera.com+1

Gates's warning matters because it breaks with the position of President Donald Trump, who has said he would not regulate the industry, and because it comes from a figure who co-founded Microsoft in 1975 and spent decades building the technology now under debate. What is new is Gates's specific claim that emergency shut-off tools are insufficient on their own, and that continuous monitoring of what models do inside laboratories is the safeguard that counts.

Why a kill switch falls short

Gates said's Meet the Press that a kill switch, an emergency cutoff for an AI system, is not enough by itself. "It's not enough to have a kill switch," he said. "I would never want to say I'm against a kill switch. But when you're trying to moderate the bad behavior, you need insight and records of what's being done." He urged Washington to mandate active tracking of models' behaviour within labs.

California Governor Gavin Newsom proposed a kill switch as part of a larger solution to AI's rate of evolution. Gates said he would never say he is against such a shut-off, but that it would not address what he called the more imminent danger. He said that a monitoring or policing layer checking whether a system is being used for a cyberattack or bioterrorism can be put in place.

Gates named bad actors using current AI tools as the most pressing threat, rather than rogue autonomous machines. "The most dangerous thing we're facing right now is people with bad intent using AI," he said. "There's never been a weapon as powerful as the combination of people with ill intent using the latest AI tools." He pointed to cyberattacks, scrambled bank accounts and shut-down power grids.

The threat Gates names first

Gates said the technology is powerful enough to drive events causing 1 billion deaths, and that preventing that could be harder than slowing the proliferation of nuclear weapons. He said regulating AI could prove more difficult than Cold War nuclear treaty negotiations, telling NBC, "This is more difficult than that was." He said small groups with AI can now do what only the biggest countries could do.

Gates dismissed the argument that safety rules would slow US technology companies down in their race against China. He said of Trump, "What he said is fine," and added that after Trump hears from him and others that regulation does not handicap the nation-state race, a consensus will come. He said that the amount of slowing down is not going to be that dramatic.

Regulation versus the US-China race

Trump has pushed back against federal intervention, writing on Truth Social earlier this month, "WHOEVER WINS AI, WINS! We are leading China, and all others, and will continue to do so." On Saturday, after meeting Chinese leader Xi Jinping to discuss AI, Trump repeated his scepticism and said he would not regulate the industry, comparing the warnings to climate change and election interference hoaxes.

Gates said he was willing to speak with Trump directly to convey his concerns, and disagreed with Trump's characterisation of the warnings as a hoax. "I hope that, given my life's work in the field, my saying how unique and different this is and how concerned I am will add to what he's hearing from other people," Gates said.

Other AI leaders have echoed the call for guardrails. Dario Amodei, chief executive of Anthropic, told the United Nations Security Council last week, "If managed poorly, I even believe AI could be a risk to humanity as a whole." Anthropic recently tapped Accenture to act as a third-party evaluator for its latest AI models, and Amodei has written that leading companies should coordinate safety standards internationally.

Democratic Representative Ro Khanna of California said that Washington and Beijing should set up a joint technical working group, and said banning recursive, self-improving AI models should not be controversial. After the summit between Trump and Xi Jinping, the White House confirmed the two nations established a working dialogue to assess AI risks, with additional talks planned.

Gates has committed resources to AI access as well as warnings: earlier this month, the Gates Foundation said it was committing $1bn to expand AI access in the developing world. The next step in the policy debate is the additional US-China talks on AI risks that the White House said are planned following the summit dialogue.