Bill Gates is done with measured optimism about AI safety. In a new essay and an interview with MIT Technology Review, the former Microsoft CEO says frontier models have already blown past the danger thresholds that experts once assumed would trigger industry-wide alarm. Bio-capabilities, cyber-capabilities, psychosocial manipulation, job-market destruction, and even basic control — all crossed, he argues, while the world mostly shrugs. “I’m just stunned at the lack of concern and discussion outside of the industry,” Gates said from his Kirkland, Washington office, describing himself as a deliberately “shrill voice” trying to force a conversation no one seems eager to have.
His bioterrorism framing is the most chilling part of the interview. Gates puts the risk of engineered pathogens at roughly 50 times that of a natural pandemic — a striking number from someone who spent years warning about exactly that scenario before COVID-19 validated those fears. “Any model that can make novel molecules should be monitored,” he said. The comparison to pandemic preparedness is instructive: the world had decades of expert warnings about coronaviruses and still fumbled the response. Gates seems to fear AI is following the same script, but faster and with fewer guardrails.
Gates isn’t only sounding alarms. He’s pushing concrete policy ideas including taxes on robots and tokens — revenue generated when AI replaces human labor — and the concept of human-reserved jobs, sectors where nations deliberately keep people employed regardless of what automation could do. The robot tax idea has been a Gates hobbyhorse for years, but pairing it with token taxes suggests he’s thinking about the full stack: physical labor displacement and knowledge-work displacement simultaneously.
There’s optimism buried in the doom, of course. Gates remains genuinely excited about AI transforming agriculture, health care, education, and even helping people navigate bureaucracy. His endgame vision is abundance. But the path there, he warns, runs through turbulence — and he’s clearly frustrated that so few people outside the AI industry seem to be looking at the weather forecast. Whether a billionaire philanthropist’s shrill voice can move policy in an election cycle remains an open question. The thresholds are already behind us, and nobody’s quite sure what that means yet.
