Anthropic whistleblower warns AI models not fully controlled

Former Anthropic researcher Jacob Coxon publicly warns that AI companies, including his former employer, do not fully control their models, citing concerns about emergent capabilities and alignment gaps. The claim is an individual expert opinion, not corroborated by internal data, but it adds to a growing chorus of insider warnings on AI safety. It matters because it could influence regulatory momentum and public trust in frontier AI firms.