Anthropic has restored Claude Fable 5 and Mythos 5 after the US government lifted an export control directive that suspended access to both models on June 12, 2026. As of July 1, Fable 5 is available again to users around the world — now guarded by a refined safety classifier that hands high-risk requests to Opus 4.8.
The Three-Week Suspension
Fable 5 and Mythos 5 launched June 9 sharing the same underlying model — Fable with mature safeguards, Mythos reserved for defensive-cybersecurity partners in Project Glasswing. On June 12, export controls arrived after Amazon researchers reported a technique that prompted Fable 5 to identify software vulnerabilities and, in one case, produce exploit code. Because the order took effect immediately and nationality could not be verified in real time, Anthropic suspended both models for all users.
What the Review Found
Anthropic’s testing found that many less capable models — including Opus 4.8, GPT-5.5, and Kimi K2.7 — could identify the same vulnerabilities and reproduce the demonstration. The episode did not expose unique Mythos-level cyber capabilities, and researchers at the US Commerce Department’s Center for AI Standards and Innovation (CAISI) validated both the prior and new safeguards as extraordinarily strong.
The New Safeguard
- Improved safety classifier blocks the reported bypass technique in over 99% of cases
- Blocked requests now fall back to Claude Opus 4.8 instead of refusing outright
- Fable 5 keeps its 1-million-token context window and 80.3% on SWE-Bench Pro
- Mythos 5 restored to select US organizations following June 26 approval, with more Glasswing partners to come
A Shared Jailbreak Framework
The episode convinced Anthropic that the industry needs a consensus standard for scoring jailbreak severity. Together with Amazon, Microsoft, Google, and other Glasswing partners, it is drafting a framework scoring capability gain, breadth, ease of weaponization, and discoverability — the basis for a common industry bar it wants codified into regulation.
What This Means
The swift resolution — three weeks from directive to redeployment — points to an emerging model for frontier AI oversight: fast government action, pre-validated industry safeguards, and a shared severity framework. For Anthropic, it also proves a capable model can return from a security scare without losing enterprise or developer trust.