Anthropic’s Claude Fable 5 is back. The US Commerce Department lifted export controls on June 30, and the model went live again on July 1 across Claude.ai, the Claude Platform, Claude Code, and Claude Cowork.
The whole thing started when Amazon researchers found a jailbreak in Fable 5 — a prompt that got the model to flag software flaws and in one case write code showing how a flaw could be abused. The feds stepped in fast. On June 12 they ordered Anthropic to cut off both Fable 5 and the less-restricted Mythos 5 for any foreign national, including the company’s own non-citizen staff. With no practical way to check everyone’s nationality in real time, Anthropic just shut both down globally.
Anthropic played down the jailbreak. They say the same tricks work on Claude Opus 4.8, OpenAI’s GPT-5.5, and China’s Kimi K2.7. The company called it routine defensive work, not some hidden super-capability.
To get the controls lifted, Anthropic built a new safety filter that blocks the specific jailbreak technique more than 99% of the time. Blocked requests get bumped down to the weaker Opus 4.8 instead. The trade-off is more false alarms on normal coding tasks.
Mythos 5 is still on a shorter leash. It came back on June 26 for roughly 100 US companies and federal agencies. Anthropic’s working to widen access.
Commerce Secretary Howard Lutnick signed off on the reversal after two weeks of review. Anthropic agreed to hunt for security problems proactively, coordinate on future launches, and report any malicious use it spots.
Former AI czar David Sacks accused Anthropic of having “prioritized the continued offering of the consumer model over safety.” University of Sydney researcher Francesco Bailo told Al Jazeera the reversal looked like the government conceding it had gone too far.
Anthropic’s also pushing for something the industry lacks: a shared scoring system for how dangerous a jailbreak really is. Four factors: capability gain, breadth, ease of weaponization, and discoverability. For the worst cases — like attacks on power grids or banks — they say they’ll deploy fixes the moment severity is confirmed.
The bigger question remains unanswered. A June 2 executive order created a voluntary review path for frontier models, but Fable 5 never went through it. When Washington needed to move fast, it reached for export controls — because there’s still no binding process, only improvised ones.
