Anthropic Brings Back Claude Fable 5 After Export Control Suspension
Anthropic has confirmed that Claude Fable 5 is officially back in action, following a tense two week suspension triggered by a US government export control order.
The saga began on June 12, when Washington moved to restrict access to Anthropic’s newest models, Claude Fable 5 and Claude Mythos 5, over concerns tied to a cybersecurity report. Amazon researchers had reportedly found a way to prompt Fable 5 into identifying software vulnerabilities, and in one instance, generating code that demonstrated how one of those vulnerabilities could be exploited. Because the directive took effect immediately and Anthropic had no real time way to confirm users’ nationalities, the company suspended access for everyone rather than risk noncompliance.
Now, with the export controls lifted as of June 30, Anthropic is rolling Claude Fable 5 back out starting today, July 1, to users across the Claude Platform, Claude.ai, Claude Code, and Claude Cowork. Pro, Max, Team, and select Enterprise plans will get Fable 5 included for up to half of their weekly usage limits through July 7, after which access shifts to usage credits. Cloud availability on AWS, Google Cloud, and Microsoft Foundry is expected to follow shortly.
Behind the scenes, Anthropic says it worked closely with government officials to build a new safety classifier specifically targeting the bypass technique flagged in the Amazon report. The company notes this fix blocks the exploit in over 99% of attempts, and that independent testers from the Department of Commerce’s Center for AI Standards and Innovation have reviewed the safeguards and called them extremely robust. Anthropic also acknowledged a tradeoff: the tighter classifier may flag more legitimate coding and debugging requests than before, something the company says it will keep refining.
Beyond the technical fix, Anthropic used the incident to push for something bigger, an industry wide standard for judging how serious an AI jailbreak actually is. Working alongside Amazon, Microsoft, Google, and other partners in its Glasswing program, the company outlined a proposed framework that scores jailbreaks on four criteria: how much capability they unlock, how broadly that capability applies, how easy the technique is to weaponize, and how discoverable it is to outsiders.
Anthropic also announced it is deepening its collaboration with the US government, including giving officials earlier access to future frontier models for independent testing, sharing threat intelligence ahead of publication, and dedicating teams and compute resources to joint AI security research.
Access to Mythos 5, the more powerful sibling model reserved for trusted cybersecurity partners, has also been restored for a set of approved US organizations following government sign off on June 26.
