An Anthropic researcher resigned while warning that major AI companies were placing competitive pressure ahead of adequate safety work. Jacob Coxon said he was resigning from Anthropic after three years of research work at Anthropic and OpenAI. He said the leading laboratories were focused heavily on outperforming one another and international competitors. Frontier AI laboratories compete for researchers, computing capacity, customers and technical leadership.
Coxon argued that the pace of development was not matched by what he considered adequate safety action. He announced the decision publicly on the social platform X. Employees may encounter confidential evidence that cannot be assessed from a short public statement. Internal safety teams can influence evaluations and deployment decisions while still disagreeing with company leadership about acceptable risk.
His departure followed other warnings from researchers inside and outside the AI industry. Anthropic has publicly emphasized safety and controllability as central parts of its company identity. Public departures can prompt scrutiny but provide less evidence than detailed technical reports or documented incidents. The reporting attributes institutional statements to the officials or organizations that made them and separates those statements from independently observed events.
The resignation is one researcher's account and does not independently establish the priorities of every team or executive. The episode arrived during a period of rapid model releases and increasing government interest in mandatory safeguards. Debate about AI governance includes voluntary standards, statutory requirements and independent evaluation. Coxon's public statement reflected his perspective, and the available report did not provide a complete internal record from either company.
Any detailed account or technical evidence Coxon releases. Changes to safety governance, evaluation gates or whistleblower protections at frontier laboratories. The next dated releases, filings or operational changes will provide the clearest evidence of whether the situation has stabilized or materially changed.