US Commerce Department partially lifts Anthropic Mythos 5 export ban for over 100 firms

The US Commerce Department has partially reversed its export ban on Anthropic’s Claude Mythos 5, the company’s most capable cybersecurity model, restoring access for more than 100 US companies and government agencies two weeks after blocking it.

The reversal came in a June 26 letter that narrows, but does not end, restrictions Commerce imposed on June 12 over concerns the model could be jailbroken for malicious use. Commerce Secretary Howard Lutnick wrote to Anthropic’s chief compute officer that ‘appropriate safeguards are in place to permit certain trusted partners to access the Claude Mythos 5 Model,’ according to the letter.

The carve-out matters because it sets an early template for how Washington gates frontier AI it considers dual-use. Anthropic said it received notice that Mythos 5, which it calls its strongest cybersecurity model, can be redeployed to a small group of cyber defenders and infrastructure providers. Fable 5, the public-facing version, remains restricted; the June 12 order had suspended access to both Fable 5 and Mythos 5, and the new letter does not mention Fable.

Anthropic’s capability claims for Mythos 5 are the company’s own and have not been independently benchmarked.

The episode reflects a tightening loop between AI labs and US export policy, with the government deciding case by case which institutions can touch the most capable security models. Whether the more than 100 cleared organizations gain access on identical terms, and how fast, will signal how durable this trusted-partner approach becomes.