Claude Fable 5 is working again. Anthropic switched it back on for everyone on July 1, 2026, after an 18-day shutdown that the company did not choose. A US export control order pulled the model offline on June 12, and it stayed dark until the Department of Commerce lifted the restriction on June 30.
If you use Claude for work, this piece walks through why the model disappeared, what Anthropic changed before flipping it back on, and what it means for your account today.
What Happened to Claude Fable 5?
Fable 5 and its sister model, Mythos 5, launched on June 9, 2026. Both run on the same base model. Fable 5 shipped with the full set of public safety guardrails, built for everyday use on Claude.ai and the API. Mythos 5 shipped with fewer of those guardrails and went straight to a small, vetted group of partners inside Anthropic’s Glasswing cybersecurity program.
Three days after launch, that changed. On June 12, the US government applied export controls to both models, citing national security. Anthropic had to cut off access “by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees”, as the company put it in its own statement.
Anthropic could not tell in real time who was a foreign national and who was not. Instead of risking a violation, it pulled both models for every user on the planet. That is the part that stung the most, since the freeze hit paying US customers just as hard as anyone else.
Why Did the Government Freeze the Models?
The trigger was a security report, not a policy debate. Researchers at Amazon found a prompting method that got Fable 5 to identify software vulnerabilities, and in one case, to produce code that demonstrated how one of them could be exploited.
That is a real concern on paper. But Anthropic pushed back on how unique the finding actually was. The company ran the same test against a stack of other models and said the results were not exclusive to Fable 5.
- Claude Haiku 4.5, Sonnet 4.6, Opus 4.6, Opus 4.7, and Opus 4.8 all reproduced the exploit demonstration.
- So did GPT-5.5 and GPT-5.4 from OpenAI, and Kimi K2.7.
- Anthropic said the technique did not touch anything unique to Mythos-level capability — it exposed a routine defensive cybersecurity task that Fable 5’s filters were blocking out of caution.
In other words: this looked less like a Fable 5 problem and more like an industry-wide blind spot that happened to get flagged on Anthropic’s model first.
When Did Access Come Back?
Fable 5 returned on Wednesday, July 1, 2026, across Claude.ai, Claude Code, Claude Cowork, and the Claude Platform. Pro, Max, Team, and select Enterprise plans get Fable 5 counted toward up to 50 percent of their weekly usage limit through July 7. After that window closes, it runs on usage credits like any other premium model call.
Cloud access is catching up more slowly. Anthropic said it will restore Fable 5 on AWS, Google Cloud, and Microsoft Foundry “as quickly as possible,” but has not given a firm date for each platform.
Mythos 5 is a separate story. The government cleared it for a limited group of US organizations on June 26, four days before the wider Fable 5 announcement. That access has not opened up beyond those vetted partners yet.
What Did Anthropic Change Before Turning It Back On?
Anthropic did not just flip a switch and hope. It built a new safety classifier targeted at the exact bypass technique from the Amazon report. In the company’s own testing, that classifier caught the technique more than 99 percent of the time.
The Commerce Department’s Center for AI Standards and Innovation reviewed both the original guardrails and the new classifier before sign-off. If Fable 5 does block a request under the new rules, that request now gets routed to Claude Opus 4.8 instead of just failing outright.
Anthropic also opened a HackerOne program so outside security researchers can report jailbreak attempts directly, and it is working with Amazon, Microsoft, and Google on a shared scoring system for how serious a given jailbreak actually is — rated on capability gain, how broad that gain is, how easy it is to weaponize, and how easy it is to discover.
Fable 5 vs. Mythos 5: What’s the Real Difference?
| Feature | Claude Fable 5 | Claude Mythos 5 |
| Availability | Public — Claude.ai, Claude Code, Claude Cowork, API | Restricted — select Glasswing partners only |
| Safety guardrails | Full public safeguards | Fewer guardrails, for defensive security work |
| Status as of July 3, 2026 | Fully restored | Restored for a limited group only |
What Does This Mean for You Right Now?
If you are on Pro, Max, or Team, log back in and Fable 5 should just be there, no re-authorization step needed. If you were mid-project when the freeze hit, this is a good week to go back and finish it while the usage limit is more generous through July 7.
If you build on the API, check that your integration is pointed at the right model string and test a few of your existing prompts against the new classifier. A handful of edge-case cybersecurity or vulnerability-research prompts that worked before June 12 may now get redirected to Opus 4.8 instead of Fable 5.
If you were counting on Mythos 5 for defensive security work and you are not already inside the Glasswing program, you are still waiting. Anthropic says it is negotiating a wider rollout, but has not committed to a date.
The Bigger Takeaway
The freeze landed at an awkward moment for Anthropic, which reportedly filed a confidential IPO prospectus showing a $47 billion revenue run rate and a $965 billion valuation right around the same time. An 18-day blackout on its newest flagship model is not the kind of headline a company wants during that stretch.
It also exposed something bigger than one company’s problem: there is no agreed-upon process for how the US government evaluates and restricts frontier AI models once they ship. A model can launch on Tuesday and get pulled by government order the following week, with no fixed review window and no clear appeals process. That is worth watching regardless of which lab it happens to next.
FAQ
Is Claude Fable 5 available right now?
Yes. Fable 5 came back online on July 1, 2026, for users on Claude.ai, Claude Code, Claude Cowork, and the Claude Platform. Pro, Max, Team, and select Enterprise plans get it included for up to half their weekly usage through July 7, after which it runs on usage credits.
What is the difference between Claude Fable 5 and Claude Mythos 5?
They run on the same underlying model. Fable 5 ships with the full set of public safety guardrails. Mythos 5 has fewer of those guardrails and is restricted to a small group of vetted organizations working through Anthropic’s Glasswing cybersecurity program.
Why did the government freeze Fable 5 and Mythos 5 in the first place?
Amazon security researchers found a prompting technique that got Fable 5 to describe how to exploit a software vulnerability. The government treated this as a national security concern and ordered Anthropic to cut off access for foreign nationals on June 12, 2026.
Why did Anthropic block everyone, not just foreign users?
Anthropic said it had no reliable way to check a user’s nationality at the moment a request came in. Rather than risk violating the order, it suspended both models for every user worldwide until it could sort out a compliant path forward.
Did other AI models have the same vulnerability?
Anthropic said its testing showed several other models could produce the same exploit demonstration, including Claude Haiku 4.5, Sonnet 4.6, Opus 4.6 through 4.8, GPT-5.4, GPT-5.5, and Kimi K2.7. The issue was not unique to Fable 5.
What did Anthropic change before turning Fable 5 back on?
Anthropic built a new safety classifier aimed at the specific bypass technique from the Amazon report. In its own testing, the classifier caught that technique more than 99 percent of the time, and the Commerce Department’s AI standards office reviewed the results before sign-off.
Is Mythos 5 available to regular users?
Not yet. Mythos 5 is currently limited to a set of vetted U.S. organizations that work on critical infrastructure defense. Anthropic is talking with the government about widening that group through the Glasswing program, but there is no public release timeline.
