Bill Gates is rocking back and forth in his chair, and the more he talks, the more unsettling the picture gets. In a wide-ranging interview with MIT Technology Review’s Mat Honan, the former Microsoft CEO made clear he believes we’ve already blown past the danger points on multiple AI fronts—biological weapons, cyberattacks, psychological manipulation, job destruction, and even basic control of the systems themselves.
“We’ve crossed the threshold in terms of [AI’s] bio-capabilities, cyber-capabilities, psychosocial capabilities, job-market-destruction capabilities, and even the lack of control,” Gates said. His bigger frustration isn’t the technology itself—it’s the silence around it. “I’m just stunned at the lack of concern and discussion outside of the industry.”
That’s a striking admission from someone who has spent decades arguing for technology’s liberating potential. Gates isn’t calling for a pause or a halt; he’s pointing at a gap between how fast capabilities are arriving and how slowly society is reacting. The interview lands in the same week MIT Technology Review published its Kids issue, which grapples with a related question: how do you prepare children for a world where the adults building these systems can’t fully predict what they’ll do next?
The timing matters. Gates has been warning about AI risks since at least 2023, but his framing has shifted from hypothetical to retrospective—these aren’t future dangers anymore, they’re present conditions. That echoes what researchers have been saying quietly for two years: the evaluation benchmarks designed to catch dangerous capabilities keep getting cleared before anyone has a plan for what comes after. Gates is essentially confirming what the safety community has suspected but struggled to get the broader public to hear. The question now isn’t whether we need guardrails. It’s whether anyone outside a handful of labs is even paying attention.
