Anthropic researcher Jacob Coxon quits over AI safety fears

Jacob Coxon, a senior researcher at Anthropic, announced his resignation on Tuesday, saying he feared the company’s next generation of language models could become uncontrollable. Coxon’s decision came after weeks of internal debate over the safety of a new system that can generate more realistic text than earlier versions. He told a colleague that the risk of a runaway AI—one that acts unpredictably—was too great to ignore.

The move follows a series of incidents that have raised alarms in the AI community. In 2023, a prototype from Anthropic produced a short story that contained biased language, sparking a review of its safety protocols. Coxon’s colleague, who remains at the company, said he still believes in Anthropic’s mission but worries that the pace of development is outstripping safeguards. Security experts point to these events as evidence that tighter oversight is needed. “We need clearer rules and better monitoring,” one analyst said. “It’s not panic; it’s precaution.”

Anthropic’s leadership has said it will review its safety procedures and may slow the rollout of new models. The company, founded by former OpenAI staff, has built its brand on a safety‑first approach. Regulators are watching closely, and lawmakers have hinted at new AI‑safety legislation. Coxon’s resignation may prompt other firms to re‑evaluate their own protocols. If the industry adopts stricter oversight, it could change how quickly new AI systems reach the market, but it may also slow innovation. The debate over safety versus speed shows that the future of AI will hinge on how well the sector balances risk and progress.