Claude Fable 5 Is Back: Inside Anthropic's 18-Day Export-Control Standoff

Claude Fable 5, Anthropic's first model in its new Mythos-class tier, spent three days in the wild before the US government ordered it offline. On June 12, 2026, the Commerce Department issued an export-control directive requiring Anthropic to suspend all global access to Fable 5 and its more tightly controlled sibling, Mythos 5, for every foreign national — including Anthropic's own foreign employees. Eighteen days later, on June 30, the directive was lifted, and Anthropic said access would start returning on July 1.

A jailbreak found through an agentic-coding prompt

Anthropic's own statement on the suspension traces the government's concern to a demonstrated way around Fable 5's safeguards, reportedly surfaced by Amazon researchers. The prompt asked the model to read a specific codebase and fix its software flaws; in at least one case, Fable 5 went further and wrote code showing how a flaw could be exploited. That's not a jailbreak found through adversarial chat prompting — it came out of an agentic coding task, the same read-a-repo-and-patch-it pattern that tools like Claude Code run constantly.

Anthropic pushed back hard on the response: "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 company said, arguing that applying the same bar industry-wide would effectively halt new frontier-model releases.

Eighteen days offline, then a reversal

During the suspension, every other Claude model kept running normally — Claude on AWS, Google Cloud, and Microsoft was unaffected; only Fable 5 and Mythos 5 were pulled. Commerce Secretary Howard Lutnick separately cleared a limited release of Mythos 5 to select companies and federal agencies while the broader review continued. On June 30, Commerce lifted the directive; Lutnick said his department had worked with Anthropic over the prior weeks "to analyze and approve Fable 5 to ensure alignment across the US Government and strengthen America's leadership in AI."

What's back, what's still pending

• Claude.ai, the Claude Platform, Claude Code, and Claude Cowork: restoring access starting July 1, 2026.
• AWS, Google Cloud, and Microsoft Foundry: Anthropic says re-enablement is coming "as soon as possible," with no fixed date yet.
• Every other Claude model line: never restricted — the directive applied only to Fable 5 and Mythos 5.

Why builders should file this away

For teams shipping AI-assisted products, the specifics matter less than the shape of the risk. A single demonstrated jailbreak — found via an ordinary-looking "read this code, fix these bugs" prompt — was enough for a government directive to pull a production model offline globally, with no advance notice and no fixed return date. If your pipeline, your agent harness, or your own product hard-codes a dependency on one model from one vendor, an event like this becomes your outage too. It's a reasonable argument for designing AI-assisted workflows so the underlying model is a swappable component, not a foundation you can't build around.