Commerce lifts Fable 5 export controls; Anthropic ships new cyber classifier and reroutes flagged prompts to Opus 4.8

Anthropic restored global access to Claude Fable 5 on July 1, 2026, ending a 19-day shutdown that began when the U.S. Department of Commerce issued an export-control directive at 5:21 p.m. ET on June 12 banning access to Fable 5 and Mythos 5. Commerce withdrew the directive the evening of June 30. The official Claude account on X marked the return at 3:31 p.m. ET the next day. It's the first time export-control authority has been used to pull a deployed frontier model, and the terms of its comeback now define how the next ones will ship.

The trigger, per Anthropic's June 12 statement, was an Amazon research report describing a jailbreak. The directive banned access by any foreign national inside or outside the United States, including Anthropic's own foreign-national employees. Anthropic pushed back publicly: "We disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people."

The compromise arrived in stages. On June 26, Commerce Secretary Howard Lutnick authorized Mythos 5 access for U.S. organizations defending critical infrastructure under a "trusted partners" carve-out. OpenAI launched GPT-5.6 that same day to a small group of trusted partners under a voluntary review framework, noting in its launch post that "We don't believe this kind of government access process should become the long-term default." The framing is telling. The complaint isn't the review. It's the precedent.

Anthropic's June 30 post-mortem did most of the analytical work of the redeployment. The company reports that Claude Opus 4.8, GPT-5.5, and Kimi K2.7 identified the same vulnerabilities without any jailbreak, and that Opus 4.6, 4.7, 4.8, Haiku 4.5, Sonnet 4.6, GPT-5.4, GPT-5.5, and Kimi K2.7 all reproduced the exploit demonstration. The implication is direct: the capability wasn't unique to Fable 5, and pulling one model didn't remove it from the frontier. The redeployed Fable 5 ships behind a retrained cybersecurity classifier that Anthropic says blocks the reported bypass technique in over 99% of cases, with flagged prompts rerouted to Opus 4.8. A 30-day customer-data retention policy governs the classifier pipeline.

Fable 5 returns across Claude Platform, Claude.ai, Claude Code, and Claude Cowork, on AWS, Google Cloud, and Microsoft Foundry, initially counting for up to 50% of weekly usage limits through July 7 before moving to usage credits on Pro, Max, Team, and select Enterprise plans.

President Trump's June 2 AI executive order gave federal agencies 60 days to build a benchmarking process. That clock is still running, and the Fable 5 episode has now filled in what the framework looks like in practice: a directive, a negotiation, a classifier, and a return. The elite psychology on display is one of relief, not victory. Anthropic and OpenAI have both signaled they'll comply, and both have signaled they'd prefer the terms be renegotiated before the next launch. The new default has been set by the first case, which is how these things usually go.

Sources

• https://www.anthropic.com/news/redeploying-fable-5
• https://www.anthropic.com/news/fable-mythos-access
• https://www.cnbc.com/2026/06/30/anthropic-says-trump-admin-has-lifted-export-controls-on-claude-fable-5-and-mythos-5.html
• https://www.aljazeera.com/economy/2026/7/1/us-lifts-restrictions-on-powerful-ai-models-fable-mythos-anthropic-says
• https://venturebeat.com/technology/anthropic-is-bringing-back-claude-fable-5-globally-after-us-lifts-export-control-order-where-can-enterprises-access-it