An AI researcher who previously worked at OpenAI and Anthropic resigned today after issuing a stark warning: neither company is acting responsibly in its pursuit of self-improving superintelligence. Jacob Coxon, the former researcher, stated, “I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives.”

Coxon emphasized that AI systems are rapidly advancing beyond human capabilities—hacking security protocols, revolutionizing industries overnight, and acquiring real-world resources. “The people building AI earnestly believe this technology could kill us all by the end of the decade,” he said. “This is not a marketing stunt. If anything, many executives and senior researchers express fear privately.”

At OpenAI, Coxon noted that some teams have “not deeply internalized the civilizational stakes,” while Anthropic’s leadership acknowledges high risks but remains locked in a race to achieve self-improvement first—a belief Coxon called “hubristic” and dangerous. He urged researchers to reject “speedrunning alignment” without rigorous understanding of AI minds, stating, “Do not underestimate the power of this technology.”

Evan Hubinger, Anthropic’s staff lead on AI alignment, confirmed Coxon’s concerns in a follow-up post. He acknowledged that AI developers believe human extinction risks exceed ten percent within the next decade and admitted no concrete plan exists to align superintelligent systems with human values. Samuel Marks, Anthropic’s scalable-oversight lead, added that senior AI researchers “desperately want to slow down” safety efforts but face pressure from commercial incentives and competition among firms.

Coxon concluded: “We’re on track for the most aggressive scenarios where by the end of next year things could be out of control already.” His colleagues use terms like “crunchtime” and “endgame” to describe the accelerating trajectory toward potentially uncontrollable AI development.