The US Department of Commerce lifted export controls on Anthropic's Claude Fable 5 and Mythos 5 models Tuesday, ending a freeze that had knocked both systems offline worldwide since June 12. Anthropic said Wednesday access would return to customers on Claude.ai, the API, and Claude Code.
"We've received notice that the Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5," the company posted on X. "We'll begin restoring access tomorrow, and will share an update soon." It's the kind of carefully bloodless statement companies write after eighteen days of not being allowed to say much of anything.
How this started
Anthropic launched Fable 5 and Mythos 5 on June 9, its first models in the new Mythos tier. Three days later the Commerce Department ordered the company to cut off access for any foreign national, including Anthropic's own non-citizen employees, citing national security authorities. Anthropic said complying with that instruction in real time was technically impossible, so it pulled both models globally rather than try to sort users by nationality on the fly.
The trigger, according to CNN reported, was a jailbreak that a trusted government partner (later identified as Amazon) found in Fable's guardrails. Anthropic pushed back at the time, calling the workaround "simple" and noting other publicly available models had similar holes. That framing didn't stop the shutdown.
What actually changed
Commerce Secretary Howard Lutnick said his office "worked closely with Anthropic to analyze and approve Fable 5 to ensure alignment across the US Government," a line that says a lot about optics and very little about mechanics. Anthropic says it built a new safeguard that blocks the specific jailbreak 99 percent of the time, a figure that comes from the company itself and hasn't been independently verified. The Center for AI Standards and Innovation reportedly tested it, but neither side has published results.
Notice also who signed the deal on Anthropic's side. Lutnick's letter to Brown, not CEO Dario Amodei, went to co-founder Tom Brown, who reportedly took over negotiations with the administration. Amodei has been an outspoken critic of light-touch AI regulation and backed Kamala Harris in 2024, which may or may not be a coincidence, but the timing is hard to ignore.
"A license is no longer required for the export, reexport, or in-country transfer, including deemed export or deemed reexport, of the Mythos or Fable models."
That's Lutnick, in the same letter, using the flattest possible bureaucratic language to describe what was, functionally, a two-and-a-half week shutdown of a commercial product over an undisclosed vulnerability.
Where things stand now
Fable 5 goes back to general availability Wednesday. Mythos 5 stays more restricted, limited to roughly 100 government-approved organizations that got access back on June 26, days before the broader Fable reversal. It's still not clear what specific technical or contractual changes tipped Commerce from ban to approval in the space of two weeks, and neither the agency nor Anthropic has spelled it out beyond the safeguard claim.
The Trump administration still faces an August deadline under its recent AI executive order to publish standardized benchmarks for evaluating frontier model risk, which is presumably where this same argument resurfaces the next time a company ships something the government hasn't seen before.
