The Trump administration has maintained export controls on Anthropic's Claude Fable 5 model after talks failed to resolve disagreements over jailbreak vulnerabilities, leaving the company's latest AI product offline.

What happened

Trump administration officials and Anthropic held talks on Monday without reaching agreement on lifting restrictions imposed on Claude Fable 5. The government remains concerned that guardrails on the model can be disabled to access more powerful cybersecurity capabilities, while Anthropic argues the administration's concerns are overblown. Commerce Secretary Howard Lutnick and Anthropic's cofounder Tom Brown and head of external affairs Sarah Heck led the discussions, with security researchers also participating.

Why it matters

This marks Anthropic's second major conflict with the US government in recent weeks—the company is already in a prolonged dispute with the Pentagon over military use of its models. The standoff signals that AI companies now face government scrutiny at launch, and the White House's willingness to restrict a model is shaping how other AI labs plan their releases. Some investors believe Anthropic may be singled out when competitors might escape similar restrictions.

What to watch

The Commerce Department expressed willingness to restore access to Fable 5 if Anthropic fully resolves the jailbreak concerns, but no timeline or specific resolution path has been announced. The outcome could set a precedent for how the US government evaluates and controls advanced AI model releases going forward.