In one of the most disruptive government-ordered AI model restrictions in history, Anthropic's most powerful publicly available AI — Claude Fable 5 — was restored to global users on July 1, 2026, after a 19-day suspension that sent shockwaves through the technology industry.
The crisis began on June 9, when Anthropic launched Fable 5 alongside its even more powerful sibling, Mythos 5. Just three days later, on June 12, the US Department of Commerce issued an emergency export control directive citing national security concerns. The trigger: Amazon researchers had discovered a jailbreak technique that could bypass Fable 5's safety classifiers, causing the model to identify software vulnerabilities and, in one case, produce code demonstrating how to exploit one.
Anthropic suspended access to both models for all users within hours — a drastic measure that affected developers, enterprises, and researchers worldwide.
What the jailbreak actually revealed
In the weeks that followed, Anthropic conducted extensive testing and found that the jailbreak did not expose unique Mythos-level capabilities. Remarkably, every model Anthropic tested — including Claude Haiku 4.5, Sonnet 4.6, Opus 4.6 through 4.8, GPT-5.4, GPT-5.5, and Kimi K2.7 — could produce the same outputs. The behavior was classified as a "borderline case" within Fable 5's deliberately large safety margin.
"The reported technique did not expose any unique Mythos-level cyber capabilities," Anthropic stated in its redeployment post. "It only involved routine defensive cybersecurity work."
New safeguards and a historic framework
To secure the model's return, Anthropic trained an improved safety classifier that blocks the specific jailbreak technique in over 99% of cases. The US Commerce Department's Center for AI Standards and Innovation independently validated the safeguards.
The trade-off: increased false positives. Some legitimate coding and debugging queries that pattern-match against the jailbreak are now blocked and rerouted to Opus 4.8, with users notified.
More significantly, Anthropic, Amazon, Microsoft, and Google agreed to co-develop a shared framework for scoring jailbreak severity — a first for the AI industry. The framework evaluates four criteria: capability gain, breadth of capability gain, ease of weaponization, and discoverability. This aims to prevent future situations where borderline findings trigger disproportionate government responses.
A new era of government collaboration
The restoration came with four formal commitments from Anthropic: pre-release government access to future frontier models for national security evaluation, rapid threat intelligence sharing, dedicated resources for joint AI security research, and participation in working toward a common industry security standard.
"Government involvement in AI releases requires a durable, transparent process that gives cyber defenders and others the certainty they need about access to powerful models," Anthropic stated. "These rules should be codified in strong regulation and applied equally across frontier model developers."
Billing cliff ahead
For now, Fable 5 is included at up to 50% of weekly usage limits for Pro, Max, Team, and select Enterprise subscribers through July 7. After that date, usage credits are required at standard API rates of $10 per million input tokens and $50 per million output tokens. The July 7 billing cliff has prompted calls from developers to prepare.
The episode has set a precedent that will likely shape how all frontier AI models are released and governed going forward — especially as the White House prepares to announce voluntary AI model standards around July 7, with an August 1 deadline for the full framework.
