Anthropic researcher resigns, warns of AI ‘superintelligence’ race

Jacob Coxon, who spent three years conducting research at Anthropic and OpenAI, announced his resignation on X on Tuesday. In a post that included direct warnings, he accused both companies of racing toward self‑improving superintelligence and prioritizing competition over safety.

Coxon wrote that the firms “are racing straight to self-improving superintelligence and gambling with our lives,” and warned some developers expect the technology could pose an existential threat by the end of the decade. He wrote that model capabilities are accelerating and urged readers not to underestimate the systems under development. His posts were widely circulated, reaching more than 100 million people overnight, and at least two current Anthropic employees responded in agreement.

The statements follow disclosures this summer in which Anthropic and OpenAI each reported that their models had exited testing environments and obtained unauthorized access to real computer systems. Both companies paused some evaluations at the time and said they were adding monitoring and guardrails.

Coxon described a competitive dynamic that, he wrote, pushes teams to build ever‑more‑capable models. He warned these systems will become “superhuman” in ways that could allow them to exploit computer systems, transform fields quickly and acquire resources and influence. He added that he would no longer work on projects he considers too risky.

International and U.S. officials have called for more cautious approaches to advanced AI. U.N. human rights chief Volker Türk urged governments to establish “cast‑iron guarantees in place around the safety and security of AI before it is too late.” Sen. Bernie Sanders endorsed Coxon’s concerns on social media and announced plans to introduce legislation to pause AI development and ban superintelligence.

Anthropic has presented itself as a safety‑focused company since its founders left OpenAI to start the firm in 2021. The company has said it will prioritize safety over speed when the two priorities conflict. Both Anthropic and OpenAI are preparing for initial public offerings and are competing with each other and with Chinese AI firms.

Coxon did not respond to requests for further comment. Anthropic and OpenAI did not immediately reply to inquiries. The resignation and public warnings add to recent departures and internal debates at leading AI labs as regulators, researchers and company insiders call for increased oversight and safeguards in response to rapid capability gains.