Fable 5 Is Back: What Changed and What to Expect

Fable 5 returned July 1 after an 18-day suspension triggered by US export controls on June 12. Learn why its safety classifiers block more requests than earlier models, what happens when they do, and how to write prompts that get through.

Fable 5 and Mythos 5 launched June 9. Three days later, on June 12, the US government applied export controls and suspended both models globally. The trigger: Amazon researchers found a method of prompting Fable 5 to identify software vulnerabilities and, in one case, produce code demonstrating how a vulnerability could be exploited. Anthropic suspended access the same day the order took effect — there was no reliable way to verify nationality in real time. Export controls were lifted June 30. Fable 5 returned July 1 for users worldwide on the Claude Platform, Claude.ai, Claude Code, and Claude Cowork. Mythos 5 came back for a set of US organizations.

Fable 5 uses safety classifiers — small automated AI systems that inspect every request for potentially harmful cybersecurity content. When a classifier triggers, the request is blocked. Fable 5 launched with a deliberately larger "safety margin" than any prior Claude model: it blocks requests that are probably benign but carry some small chance of harm. The result is more false positives — routine security work may get flagged. When Fable 5 blocks your request, you are notified and the same request is automatically sent to Opus 4.8. You do not need to retype or switch models manually.

Anthropic, Amazon, Microsoft, and Google are building a shared framework to score jailbreak severity. Every jailbreak is rated on four factors: Capability gain (does the technique unlock something other models already do?), Breadth (does one technique unlock many harmful behaviors or just one?), Ease of weaponization (does it work on the first try, or require skilled effort?), and Discoverability (is it already public, or specialist knowledge only?). The Amazon-found bypass scored low on all four — Anthropic confirmed every model tested, including Claude Haiku 4.5 and GPT-5.5, could produce the same output. That finding was central to why export controls were lifted.

Fable 5 is available on Pro, Max, Team, and select Enterprise plans. Through July 7, those plans include Fable 5 for up to 50% of weekly usage limits. After July 7, Fable 5 requires usage credits. Standard Enterprise plans have no included Fable 5 allowance — credits only. Premium Enterprise plans had Fable 5 included through July 7; enabling credits after that keeps access on. AWS, Google Cloud, and Microsoft Foundry access is being restored as quickly as possible. Anthropic has also launched a HackerOne program where security researchers can submit potential Fable 5 cyber jailbreaks for review.

Fable 5 returned July 1 after an 18-day suspension. Its safety classifiers are set wider than any prior Claude model — expect blocks on security-adjacent work, with automatic fallback to Opus 4.8. Framing prompts around remediation rather than exploitation is the most reliable way to get useful output from either model. The 4-factor jailbreak severity framework (capability gain, breadth, ease of weaponization, discoverability) is now being standardized across Anthropic and major cloud partners. After July 7, Fable 5 requires usage credits on most plans.