Fable 5 Returns Globally as US Lifts Export Controls on Anthropic
Anthropic restored global access to Claude Fable 5 on July 1 after the US Commerce Department lifted the 19-day export controls. Claude Code and Claude.ai users get the model back with a new 99%+ jailbreak classifier and an industry-wide safety framework now in development.
Anthropic
Fable 5 Returns to Claude Code and the API After a 19-Day Suspension
Anthropic restored global access to Claude Fable 5 on Wednesday, July 1, after the US Department of Commerce lifted the export controls that had kept the model offline since June 12. Access is now live on Claude Platform, Claude.ai, Claude Code, and Claude Cowork. AWS, Google Cloud, and Microsoft Foundry re-enablement is underway.
For Pro, Max, Team, and select Enterprise subscribers, Fable 5 counts toward up to 50% of weekly usage limits through July 7 at no extra cost. After July 7, the model is available via usage credits. Standard Enterprise seats require credits to be enabled.
The suspension started three days after Fable 5's June 9 launch, when the US government became aware of an Amazon research report describing a prompt that bypassed the model's cybersecurity safeguards. The jailbreak caused Fable 5 to identify software vulnerabilities and, in one case, produce code demonstrating how a flaw could be exploited.
What Changed With the Safeguards
Anthropic deployed a new safety classifier built in collaboration with the US government and Amazon. The classifier targets and blocks the reported bypass technique in over 99% of cases. When a request is flagged, it is automatically rerouted to Claude Opus 4.8 rather than returning an error.
The tradeoff is an increase in false positives during routine coding and debugging. Anthropic acknowledged this and said it will continue refining the classifier to reduce unnecessary blocking of legitimate developer requests.
Anthropic's own testing confirmed that the jailbreak did not expose unique Mythos-level offensive capabilities. The same vulnerabilities could be identified by Claude Haiku 4.5, Sonnet 4.6, Opus 4.8, GPT-5.4, GPT-5.5, and Kimi K2.7 using an identical prompt. Researchers from the US Department of Commerce's Center for AI Standards and Innovation tested both the prior and updated safeguards and described them as "extraordinarily strong."
The Industry Jailbreak Framework
The most consequential outcome of the suspension may be the framework it triggered. Anthropic is partnering with Amazon, Microsoft, Google, and other Glasswing partners to develop a consensus scoring system for AI jailbreak severity — something the industry currently lacks.
The proposed framework scores each jailbreak on four criteria: capability gain, breadth of capability gain, ease of weaponization, and discoverability. Anthropic plans to calibrate its response based on severity, with 24/7 monitoring and immediate remediation for the most serious findings. A HackerOne program is now live for security researchers to submit Fable 5 jailbreak reports.
Mythos 5 Remains Restricted
Claude Mythos 5 — the same underlying model as Fable 5 but with fewer safety restrictions — remains available only to approved US organizations under Project Glasswing. Anthropic said it is working with the government to expand access to a broader set of domestic and international Glasswing partners.
Deeper Government Collaboration
In exchange for restored access, Anthropic committed to four areas of expanded government collaboration: pre-release access and evaluation for capability-frontier models, rapid information sharing on jailbreaks and safeguards, dedicated joint research resources, and participation in developing a voluntary industry evaluation standard.
Commerce Secretary Howard Lutnick said the department coordinated with Anthropic after the company agreed to "proactively detect and address security risks" and to report any malicious activity it finds in its models.
What Developers Should Know
Claude Fable 5 is priced at $10 per million input tokens and $50 per million output tokens, with existing 90% input token discounts for prompt caching. The API model string is claude-fable-5.
Developers should expect some additional false positives in cybersecurity-adjacent debugging and analysis tasks under the strengthened classifier. For teams with production pipelines built on Fable 5, Monday's episode established a clear precedent: a frontier model can be pulled offline by regulatory action with hours' notice. Maintaining fallback paths to Opus 4.8 is no longer a nice-to-have.
