US lifts export controls on Anthropic's Fable 5 and Mythos 5 three weeks after blocking them

Anthropic confirmed that the Department of Commerce has lifted export controls on its Fable 5 and Mythos 5 frontier models, roughly three weeks after they were blocked following a jailbreak discovered days after their June 9 launch. In a post that drew 83,903 likes and 12,995 retweets on X, the company said it had received notice of the reversal and would 'begin restoring access tomorrow,' thanking users for their patience and everyone who worked on redeployment.

The episode is significant less for the models themselves than for what it reveals about frontier-AI governance: the US government now demonstrably intervenes in the release and export of specific commercial model versions, coordinating with labs during security reviews before restoring access. That same mechanism is visible in OpenAI's parallel restriction of GPT-5.6 Sol to government-approved customers, suggesting a durable new pattern rather than a one-off.

The developer and research community reacted with cautious optimism mixed with unease. On Hacker News, a thread on the lifted controls drew 923 points and 641 comments, with commenters calling the government-coordination model a governance precedent while worrying about 'impossible standards' like zero-jailbreak guarantees. Wharton's Ethan Mollick publicly called for an official government statement about the risks seen in Fable, how defenders should prepare for coming open-weight Mythos-class models, and whether the worry is state actors or independent hackers.

The reversal coincides with Anthropic's broader product push — Claude Sonnet 5 and the Claude Science workbench — and with its escalating public dispute with Alibaba over an alleged distillation attack. What to watch: whether Commerce publishes any formal rationale, and whether the restore-access timeline slips or comes with new conditions on frontier deployment.