Amazon CEO reportedly raised concerns before Anthropic restrictions

According to The Wall Street Journal, Amazon CEO Andy Jassy raised technical security concerns with White House officials about Anthropic's most advanced models.

Reporting by TechCrunch and TechPolicy Press says the U.S. Department of Commerce issued an export-control directive that led Anthropic to disable access to `Fable 5` and `Mythos 5` for foreign nationals.

Anthropic said in a Friday blog post that the government "believes it has become aware of a method of bypassing, or 'jailbreaking' `Fable 5'", and the company disabled access to both models for all customers to satisfy the directive, per TechPolicy Press and TechCrunch.

MindStudio reported that Mythos previously scored 83.1% on cybersecurity evaluations versus 66.6% for a comparator model, and that Anthropic had been running a controlled, limited rollout prior to this intervention. These reported evaluation results help explain why observers and some government officials flagged cybersecurity risk as the salient concern.

Industry reporting frames this episode as a rare instance of informal government intervention in model rollouts, with TechPolicy Press describing the Commerce action as one of the most powerful examples yet of U.S. government intervention in the AI race. TechPolicy Press also quoted former Trump administration AI advisor Dean Ball calling the move "baffling."

Companies developing frontier models face a growing operational risk from export controls and ad hoc government requests, according to public coverage. Independent security testing and documented mitigation steps are increasingly likely to be focal points for both policymakers and enterprise customers, as illustrated by the attention given to alleged jailbreak methods in this case.

Observers should track official statements or filings from the Bureau of Industry and Security at the Department of Commerce for clarifying language on the legal basis and scope of the directive, and any technical disclosures from Anthropic about the alleged bypass. Also monitor follow-on reporting for named-actor security research that demonstrates or refutes the claimed jailbreak, and any shifts in how cloud providers and partners handle access controls for advanced models.