Jacob Coxon, a 27-year-old researcher, resigned from Anthropic on September 8, 2026. He previously spent three years working on frontier models at both OpenAI and Anthropic. In a public statement on X, he accused these companies of prioritizing development speed over essential safety assessments.
On September 13, 2026, Coxon appeared on NBC News’ “Meet the Press” to advocate for global coordination between AI labs and government bodies. He warned that while current models do not pose an immediate threat, risks could escalate as capabilities scale over the next 6 to 12 months.
The researcher highlighted concerns that catastrophic outcomes could emerge by the end of the decade. His claims are supported by comments from Evan Hubinger, Anthropic’s alignment science lead, who acknowledged the lack of a clear strategy for superintelligence alignment and estimated a greater than 10% probability of human eradication within a decade.
The resignation coincides with Anthropic’s preparations for an IPO, which could potentially value the company at up to $2 trillion. Coxon’s departure follows a series of similar exits from major AI labs, including recent departures from OpenAI related to safety disagreements.
