Commerce Secretary Howard Lutnick told Anthropic in a letter Friday that the company had made "significant progress" addressing government concerns and that "appropriate safeguards" were now in place for limited release.
The Trump administration ordered Anthropic on June 12 to suspend access to Mythos 5 and Fable 5 by any foreign national, inside or outside the US, days after the models launched. The government never publicly specified its concerns, though Anthropic had previously acknowledged Mythos 5 was powerful enough to detect code exploits threatening software security.
Anthropic confirmed on X it is restoring access for "a set of US organisations that operate and defend critical infrastructure" and continuing talks with the government "to expand access to Mythos 5 and make Fable 5 available for general use again." The Commerce Department has not published the list of approved organisations.
Analysis
This is selective re-opening, not a return to normal. The US government is now operating a tiered-access framework for frontier models — approved American entities on one side, everyone else on the other. OpenAI separately acknowledged government requests to delay GPT-5.6's full public release. If every frontier model now ships with a government-vetted whitelist, the open-access era of AI is ending. The question is whether this framework expands — or tightens further.
