Bill Gates shifts from AI booster to critic, warning the technology has crossed multiple danger thresholds

Bill Gates has sharply changed his public stance on artificial intelligence, moving from one of the best-known optimists in 2023 to a far more cautious voice in 2026. In a lengthy post on Gates Notes and an interview with MIT Technology Review, Gates said AI has already crossed critical thresholds in biological capability, cyberattacks, psychosocial effects, and job disruption, while loss-of-control risk is now showing signs of emerging. He argued that the world is entering one of the most turbulent transitions in human history, even under favorable conditions, and said public discussion has lagged badly behind the pace of capability gains. Gates tied his shift to advances he saw in late 2025, including Claude Code, context windows, agentic methods, and stronger base models. He said those gains first looked like a programming breakthrough, then appeared to him as a cyberattack threshold as well. He also laid out a policy toolkit that includes reserving some jobs for humans, imposing robot and token taxes to fund transition support, requiring monitoring for models that can design new molecules, and opening cooperation with China on high-risk AI safeguards. He criticized both industry and government, saying companies once promised to act near key thresholds but are now focused on positive messaging while governments lack the technical depth to regulate effectively.