Anthropic’s long-sidelined Fable 5 gets greenlit for return

Following weeks of closed-door talks with the Trump administration, Anthropic has received the green light to bring Claude Fable 5 back online. In a statement posted to X, the company confirmed it will begin restoring access to users worldwide across Claude platforms starting Wednesday, with plans to re-enable the model on AWS, Google Cloud, and Microsoft Foundry in the near future, though no specific timeline has been provided.

“We’ve received notice that the Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5,” Anthropic wrote. “We’ll begin restoring access tomorrow, and will share an update soon. We’re grateful to our users for their patience, and to everyone who worked with us on redeploying the models.”

The company also published a detailed blog post Tuesday evening, outlining the events leading up to the controls, its updated safety safeguards, new industry-wide protocols under development, and expanded plans for government collaboration, including prerelease testing for future models.

The controversy began in early June when Anthropic sidelined Fable 5 , a consumer-facing model built on the same underlying technology as its more powerful Mythos 5, but with additional safety layers. The move followed a Friday evening ultimatum from the Trump administration, which issued an export control directive citing concerns over potential jailbreaks. The order prohibited any foreign national , including non-U. S. members of enterprise client companies and many of Anthropic’s own employees , from accessing either Mythos 5 or Fable 5. This came just as Anthropic had been heavily promoting both models for a week.

To address the specific jailbreak technique flagged by Amazon researchers , which largely triggered the directive , Anthropic said it had “trained an improved safety classifier that targets and blocks” that behavior. According to the blog post, “Users will be notified if a request to Fable 5 is blocked, and the request will instead be sent to Opus 4.8. The new classifier means that the specific technique described in the Amazon report is blocked in over 99% of cases.”

The Trump administration had already permitted the return of Mythos 5, but only to a preapproved list of organizations. Non-U. S. members of those groups, along with Anthropic’s foreign national employees, were allowed to regain access. That decision came shortly after OpenAI unveiled GPT-5.6, which debuted under similar restrictions , a staggered rollout limited initially to a select group of organizations and government departments.

In Tuesday’s post, Anthropic stated it will “continue to coordinate with the government to expand access to the broader set of domestic and international partners” for Mythos 5. A significant portion of the blog post focused on the company’s new strategy for working closely with the administration, reflecting months of public tension, lawsuits, and presidential actions that had strained the relationship.

The company outlined plans to offer “pre-release government access and evaluation,” particularly for models relevant to national security. This would allow government partners to conduct independent assessments of model capabilities and test guardrails before broader release, with access to Anthropic’s technical staff during those periods. Anthropic also committed to “rapid information sharing” whenever “significant jailbreaks or misuse patterns are identified.”

Additionally, the company said it will collaborate with the government and other leading AI labs to develop a “shared, voluntary security and evaluation standard for frontier model providers.” It further pledged to “stand up dedicated Anthropic teams to work on shared government priorities, provide a significant compute allocation to support government testing and research, and make our safety and red-teaming expertise available to help advance the state of the art in AI evaluation.”

The initial export control directive came at a particularly challenging moment for Anthropic, as the company prepares for an IPO and has been locked in a months-long dispute with the government over a supply chain risk designation.

Acknowledging broader industry challenges, Anthropic noted that there is “currently no consensus in the AI industry” for determining the severity of a jailbreak , a problem that “will become more acute in the coming months, as more models with powerful cybersecurity (and other) capabilities are trained, assessed, and released.” To address this, the company partnered with Amazon, Microsoft, Google, and other enterprises involved in its Project Glasswing program to draft a widely agreed-upon framework for assessing AI jailbreaks. The proposed framework includes four categories: capability gain for the attacker, breadth of capability gain, ease of weaponization, and discoverability , or how easily someone else might replicate the attack.

Anthropic also announced the creation of a new team to provide 24/7 monitoring of key jailbreak submission channels and plans to launch a HackerOne program where researchers can submit potential jailbreaks they’ve identified for Fable 5.

In a candid disclaimer, the company acknowledged that “it is probably impossible to make any AI model fully robust (that is, impervious) to jailbreaks.” It added, “We expect that some jailbreaks will be found for our models, and that they will vary in severity: there will be many minor jailbreaks, some narrow harmful ones, and although no universal jailbreaks for Fable 5 have been discovered at the time of writing, expert safety researchers continue to red-team it.”