An AI researcher has resigned from Anthropic, citing concerns that leading AI labs are prioritizing speed over safety in the development of advanced systems. Jacob Coxon, who previously worked at OpenAI and Anthropic, announced his departure on X (formerly Twitter) on Tuesday, stating that neither company is acting responsibly in the pursuit of self-improving superintelligence.
Coxon, who spent three years conducting pre-training research at both companies, wrote that executives and senior researchers privately acknowledge existential risks posed by AI. "The people building AI earnestly believe that it could kill us all by the end of the decade," he stated. "This is not a marketing stunt." He added that while OpenAI’s staff may not fully grasp the "civilizational stakes," Anthropic’s leadership understands the risks but feels compelled to race ahead to prevent others from doing so.
Coxon’s resignation follows a public response from Evan Hubinger, an Alignment Science Lead at Anthropic, who agreed with his assessment. Hubinger stated that he personally believes there is a greater than 10% chance AI could cause human extinction within the next decade, though he noted that current AI models pose relatively low risk. His concerns center on future superintelligent systems capable of recursive self-improvement, a scenario Anthropic has previously discussed in its safety research.
Background and Context
Coxon joined OpenAI in 2023, where he contributed to research on GPT-4o before transitioning to Anthropic in July 2026. His departure note highlighted a lack of transparency from both companies regarding their internal assessments of AI risks. He argued that the industry’s competitive dynamics are driving reckless behavior, with firms prioritizing rapid advancement over safeguards.
Anthropic, known for its focus on AI safety and alignment, has long emphasized the importance of responsible development. However, Coxon’s resignation suggests that even within safety-conscious organizations, concerns persist about the broader industry’s trajectory. Hubinger’s public endorsement of Coxon’s claims underscores the depth of these internal debates.
Industry Response and Ongoing Debates
The resignation has reignited discussions about AI governance and the balance between innovation and risk mitigation. Critics of the current approach argue that the race for advanced AI lacks sufficient oversight, while proponents contend that self-regulation and industry-led safety measures are the most effective path forward. The debate reflects broader uncertainties about how to manage technologies with existential potential.
Coxon’s departure does not indicate an immediate policy shift but adds to a growing chorus of warnings from AI researchers about the unchecked acceleration of AI capabilities. The incident highlights the tension between competitive pressures and the ethical imperative to address catastrophic risks.