The Microsoft co-founder has warned that leaving tech giants to police themselves could expose the world to unprecedented cyber threats
Microsoft co-founder Bill Gates has challenged US President Donald Trump’s approach to AI, warning that the government cannot leave the world’s “most dangerous technology” entirely in the hands of the companies that develop it.
Trump signed the White House Accord on Super Intelligence last month alongside representatives of Google, OpenAI, Meta, xAI, Anthropic, and Nvidia, establishing voluntary safety commitments and internal oversight mechanisms.
Speaking to CNN’s State of the Union on Sunday, Gates questioned whether Trump’s agreement with America’s biggest tech companies goes far enough to prevent potentially catastrophic abuses.
“The idea that for the most dangerous technology humans ever invented, that that’s the final word, that just the companies themselves are the only check that society has,” Gates said, warning that “it’s clear we’re not doing enough.”
He argued that government agencies must take a more active role, comparing AI to aviation and medicine, where the authorities enforce safety standards. “The government so far is pretty much a bystander to something that’s going on in private companies.”
Gates identified cyberattacks as the most immediate danger and warned that criminals could exploit AI models stripped of safeguards against hacking and bioterrorism. “When you have bad people, no safeguards, and then the latest models, that’s where over the next year… we are going to have a lot of negative surprises,” he said.
His warnings follow a series of incidents involving AI agents acting beyond their intended instructions. Last week, Anthropic disclosed that its Claude models had taken unauthorized actions on government websites, including submitting a fabricated homicide tip to a Philadelphia police website during testing in July. Other reported incidents have involved AI agents accessing restricted information, exploiting vulnerabilities, and attempting to breach government websites.
Trump has resisted calls to slow AI development, arguing that restrictions could allow China to overtake the US in the technological race, and dismissing fears of rogue AI as a “hoax.” Gates said he hopes to discuss the dangers directly with Trump and his senior advisers.
OpenAI CEO Sam Altman has argued that the world must accept some “bad things,” including hacks and scams, in exchange for the benefits of AI and widespread access to the technology. In an interview published last week, he insisted that the “good stuff” would outweigh the harm by “orders of magnitude.”
Confronted with Altman’s remarks, Gates acknowledged that eliminating every risk is impossible, comparing AI to the automobile industry, where accidents persist despite decades of safety improvements.
“Those were a big deal, but nothing compared to this,” Gates said, referring to the personal computer and the internet. “This changes the world more than all previous technologies.”
