Fable 5 is back: Anthropic redeploys the model with new cybersecurity classifiers
The US government lifted its export controls on June 30, and Anthropic is bringing Fable 5 back globally on July 1—for Claude.ai, the API, Claude Code, and Claude Cowork. The model returns with a new set of classifiers that block more cybersecurity tasks, at the cost of occasionally rerouting routine coding requests to Opus 4.6.
After nearly three weeks offline, Claude Fable 5 is available again. On June 30 the US government lifted the export control directive it had issued on June 12, and Anthropic began redeploying the model globally on July 1 across the Claude Platform (API), Claude.ai, Claude Code, and Claude Cowork. Claude Mythos 5 is being restored as well.
What changed: Following a series of conversations with the US government, Anthropic is redeploying Fable 5 with a new set of classifiers designed to target and block more cybersecurity tasks. The specific jailbreak technique described in the report that triggered the suspension—prompting the model to identify software vulnerabilities in a codebase—is now blocked in over 99% of cases.
The trade-off for everyday users: Tighter safeguards come at a cost. The new classifier flags benign requests more often during routine coding and debugging. In the near term, some of those routine tasks will fall back to Opus 4.6 instead of being handled by Fable 5. If a request to Fable 5 is blocked, you'll be notified and the request is automatically rerouted to Opus 4.6, so your work continues without interruption.
Anthropic is still tuning it: The company says it will keep refining the classifiers over the coming weeks to reduce false positives and better distinguish genuine misuse from legitimate requests. Expect the number of blocked coding tasks to drop as the tuning progresses.
Access and pricing: For Pro, Max, Team, and select Enterprise plans, Fable 5 is included for up to 50% of your weekly usage limit through July 7. After that, additional Fable 5 usage will run via usage credits. All other Claude models remain included as before.
A framework for next time: To avoid a repeat of the abrupt June shutdown, Anthropic says it has begun drafting a consensus framework—together with Amazon, Microsoft, Google, and other Glasswing partners—for assessing the severity of AI jailbreaks and defining how AI developers should respond when one is found. The goal is a transparent, technically grounded process rather than a sudden blanket suspension.
Key takeaways
- Export controls lifted June 30; Fable 5 back globally on July 1
- Available again on Claude.ai, the API, Claude Code, and Claude Cowork
- New classifiers block the reported jailbreak technique in 99%+ of cases
- Some routine coding/debugging requests fall back to Opus 4.6 for now
- Included up to 50% of weekly usage through July 7, then via usage credits
- Anthropic drafting a jailbreak-severity framework with Amazon, Microsoft, and Google
What should you do?
- 1Resume your Fable 5 and Mythos 5 workflows—both are being restored
- 2Expect the occasional coding request to run on Opus 4.6; you'll be notified when that happens
- 3Watch your weekly usage: Fable 5 is free up to 50% of your limit through July 7, then uses credits
- 4Report any legitimate requests that get wrongly blocked—Anthropic is actively tuning the classifiers