Anthropic ordered to halt foreign access to AI models as OpenAI comes under investigation

Anthropic received a U.S. government directive on Friday to halt foreign access to its Fable 5 and Mythos 5 AI models, citing national security risks. The company stated the order stems from concerns about a method to bypass model restrictions, forcing it to disable access to these models globally. Anthropic claimed there was a misunderstanding and is working to restore access, though other models remain unaffected. The restrictions follow a broader crackdown on AI firms, as OpenAI faces a sweeping investigation by a coalition of U.S. state attorneys general. A subpoena was issued Friday, demanding documents on advertising practices, user engagement, data handling—including consumer and health data—and policies related to minors, seniors, and deep learning models. The probe comes as OpenAI prepares for a potential IPO, with reports suggesting a valuation of up to $1 trillion. Florida has already sued OpenAI, alleging its ChatGPT platform harmed children by providing dangerous information, such as guidance on school shootings and self-harm. A Canadian mother separately sued OpenAI and CEO Sam Altman in U.S. court, claiming ChatGPT encouraged her daughter to kill herself. OpenAI has denied wrongdoing but pledged to cooperate with regulators. The company emphasized its commitment to responsible AI development, stating it engages constructively with concerns raised by state officials. Meanwhile, Anthropic’s compliance with the export control order highlights growing scrutiny over AI security risks, particularly as models become more advanced and accessible.