The AI Could Be out of Control by Next Year Says the Researcher Who Just Quit Over It

Photo Credit: Anthropic

"Neither company is acting responsibly," Coxon wrote. "They are racing straight to self-improving superintelligence and gambling with our lives." Speaking to the Wall Street Journal, he said the field is "on track for a lot of the most aggressive of these scenarios where by the end of next year things could be out of control already."

His fear centres on recursive self-improvement, the point at which AI takes over much of the work of building better AI, designing and training smarter successors in a loop that speeds up faster than humans can supervise. Once that begins, Coxon argues, the systems could improve so fast that they stop following human instructions.

What makes the exit notable is that Coxon joined Anthropic specifically for its safety reputation, and still calls that work sincere. His point is that good intentions cannot survive a race, and no lab can stay safe while its rivals push ahead. Colleagues, he said, now use words like "crunchtime" and "endgame" in everyday talk.

The timing is awkward for Anthropic, which is preparing for a public listing even as it warns that powerful AI may eventually need to be slowed. On the same day, OpenAI chief scientist Jakub Pachocki published an essay describing modern AI as an "alien mind" and calling for global coordination.

Many researchers consider such timelines far too aggressive, and Anthropic has not answered Coxon's specific claims. Still, the warning is hard to wave off, coming from someone who was doing the work and chose to walk. By his own reckoning, the answer is not far away.