U.S. bars foreign access to Anthropic's top AI models; company disables them

What This is

• The U.S. government ordered Anthropic to block foreign nationals from accessing its newest, most advanced AI models, prompting Anthropic to disable those models entirely while it works to comply.
• The restriction targets Anthropic’s Claude Fable 5 and Claude Mythos 5 models, and the directive’s duration and path to restoring access have not been made clear in the reporting.
• The U.S. action was framed as an export-control and national-security measure, with officials pointing to concerns that the models’ safeguards could be bypassed via a “jailbreak” technique.
• Anthropic says the government’s concern centers on a narrow, non-universal jailbreak involving minor and previously known issues, and the company disputes that taking the models offline is warranted.
• The order applies to “foreign persons,” and reporting says it also bars Anthropic’s own foreign-national employees from accessing the affected models.
• Commerce Department export-control rules are being used to require licenses for transfers outside the U.S. or to foreign persons, and Anthropic halted broader distribution while assessing compliance.
• Other Anthropic models are not expected to be affected, and the company says it is working to restore access to Fable 5 and Mythos 5.

Why This Matters

• It tests how far export controls can reach when “access” to software is treated like a cross-border transfer.
• Blocking foreign-national employees can reshape hiring, research collaboration, and where frontier AI work happens.
• The episode accelerates a shift toward access-tiered AI, where a few vetted users get stronger capabilities than the general public.
• It raises governance questions about due process and transparency when national-security agencies intervene in private tech operations.