The US has removed export restrictions on Anthropic’s Fable 5, allowing the model to return.

      The US Commerce Department has removed the export restrictions it had imposed on Anthropic’s most advanced models, concluding a nearly three-week suspension that had taken the company’s Claude Fable 5 offline. Reuters first disclosed this development on Tuesday, June 30, citing an individual familiar with the situation. Shortly thereafter, Anthropic confirmed the news in a post titled “Redeploying Claude Fable 5.”

      The controls had affected both Fable 5, which is publicly accessible, and Mythos 5, the more advanced cybersecurity system that underpins it. Both models received clearance simultaneously. Commerce Secretary Howard Lutnick announced that an export license would no longer be necessary. In return, Anthropic committed to actively identify and mitigate security risks, assist in the establishment of standards for future models, and report any malicious activities to the government.

      The restrictions had been implemented around June 12 and were reversed roughly two and a half weeks later. The prior order required Anthropic to swiftly discontinue both Fable 5 and Mythos 5. The catalyst for this action was a jailbreak; researchers, reportedly from Amazon, discovered a method to circumvent Fable 5’s safeguards and access the underlying capabilities of the Mythos model.

      Throughout the situation, Anthropic challenged the seriousness of the security flaw, contending that it did not justify a total shutdown. Regulators initially took a firmer stance. It is important to note the distinction between the two systems: Mythos 5 is the potent cybersecurity model, while Fable 5 serves as the consumer-facing interface built on top of it, which is why a jailbreak of one raised concerns regarding the other.

      The original restrictions may be connected to worries that a group associated with China could exploit the vulnerability to access Mythos. This reasoning is based on anonymous sources and should be regarded as reported rather than confirmed. The government's stance seems to have softened in subsequent weeks. Various reports suggest that the reversal was influenced by industry criticism and concerns that international competitors could gain an advantage while a leading US model remained inactive, although this interpretation is more analytical than official.

      When will it come back and in what manner? Anthropic stated that Fable 5 would be reinstated globally on July 1 across its offerings, including the Claude platform, Claude.ai, Claude Code, and Claude Cowork. The rollout will be intentionally slow, with usage limited to up to 50% of normal weekly limits until July 7, after which full access will be restored, according to the company's statement.

      This three-week episode has been notably public for an export-control dispute, which typically unfolds quietly through the Bureau of Industry and Security. In this case, the model was named, and the company provided real-time updates on its suspension and subsequent restoration.

      This situation occurs amidst a broader tightening of US policies regarding the export of advanced AI technologies. Over the past year, Washington has been closing loopholes that allowed state-of-the-art chips and models to reach Chinese firms via overseas subsidiaries.

      For Anthropic, the immediate stakes encompassed both commercial and regulatory aspects. A stalled flagship model could drive paying customers to competitors, creating an increasingly significant competitive gap. The company had also spent a considerable portion of June in discussions in Washington, advocating its position as the freeze persisted. The conditions imposed by Lutnick for lifting the restrictions seem to reflect the price of resuming operations.

      Anthropic had received approval to reinstate Mythos 5 for a group of trusted cyber-defense partners prior to this week, while Fable 5 remained restricted. Tuesday’s decision addresses that remaining limitation.

      What is clearly established is straightforward: the Commerce Department lifted the controls, Lutnick outlined conditions, Anthropic accepted them, and access is being restored. The political motivations behind the reversal, including the specific role of the alleged China connection and the weight of competitive pressure in the final decision, remain less clear. Regarding these matters, the sources are less robust, and a more comprehensive narrative is likely to unfold in the coming weeks.