Claude Fable 5 is back: Anthropic ends its 19-day export-control suspension with a new cybersecurity classifier
Anthropic redeployed Claude Fable 5 on 1 July 2026 after a 19-day suspension triggered by a jailbreak that let the model identify software vulnerabilities, adding a classifier that blocks the technique in over 99% of cases plus a new Cyber Jailbreak Severity framework.
4 July 2026
Claude Fable 5 and Mythos 5 launched on 9 June 2026. Three days later, on 12 June, Anthropic suspended them under US export controls after Amazon researchers found a jailbreak that let Fable 5 bypass safeguards and identify software vulnerabilities — a capability serious enough to trigger national-security restrictions. On 1 July, after a 19-day suspension, Anthropic redeployed both models globally.
What changed to make redeployment possible
Anthropic didn’t just patch the specific jailbreak — it added defense in depth. A new classifier blocks the reported technique in over 99% of cases and automatically reroutes flagged requests to Claude Opus 4.8 instead. Anthropic also accepted government conditions on deployment, opened a HackerOne program so security researchers can report cyber jailbreaks directly, and proposed a Cyber Jailbreak Severity (CJS) framework intended to standardise how AI jailbreak risk gets measured across the industry — an attempt to get ahead of the next model that ships with similar capability.
Fable 5 is now available again across the Claude Platform, Claude.ai, Claude Code, and Claude Cowork for Pro, Max, Team, and select Enterprise plans.
Why this matters beyond the headline
The pattern is becoming a template: frontier models are increasingly capable of things like autonomous vulnerability discovery, and export control review is now a real step in the release pipeline, not a formality. A 19-day gap between “most capable model we’ve shipped” and “back online with a public bug-bounty channel” is fast by regulatory standards, but it’s still 19 days where teams building on that model had to plan around unavailability.
So what
If your product roadmap depends on a specific frontier model’s capabilities, build in the assumption that safety- or export-related suspensions can happen and won’t always be brief. That means designing model-agnostic integration layers where practical, and treating vendor announcements about safeguards and severity frameworks as signal about what capabilities are coming next, not just compliance news. We factor this kind of platform risk into how we scope AI-assisted builds — more on that on the AI-assisted development page, or get in touch if you’re weighing which model or vendor to build around.