Claude Fable 5 Now Available Globally with Enhanced Safety Measures

Anthropologic has re-deployed Claude Fable 5 globally and is now available for all users as of July 2, 2026. The rollout includes Anthropic’s new safety classifier for Claude Fable 5, which blocks the specific jailbreak technique that Amazon reported in more than 99% of cases.

Cyber Jailbreak Severity Framework Introduced

Anthropric has proposed an AI jailbreak severity framework called the Cyber Jailbreak Severity (CJS) scale, developed in partnership with Amazon, Microsoft, Google, and other Glasswing partners. The CJS scale rates jailbreak severity from CJS-0 (Informational) to CJS-4 (Critical) using exponential rather than linear bands, with each step up intended to be several times more serious than the last.

Safety Classification System

Claude Fable 5’s safety classifiers categorize cybersecurity activities into four distinct risk tiers: prohibited use, high-risk dual use, low-risk dual use, and benign use. When blocked requests are submitted to Claude Fable 5’s classifiers, they are automatically rerouted to Claude Opus 4.8 and the user is notified.

Bug Bounty Program Launched

Anthropic launched a HackerOne bug bounty program where security researchers can submit potential cyber jailbreaks discovered in Fable 5 for review.

Source: Anthropic