Anthropic restores global access to Claude Fable 5 after patching Amazon-discovered exploitation bypass, new classifier blocks technique in over 99% of cases

Anthropic restored global access to Claude Fable 5 on July 1, 2026, ending an 18-day suspension that followed US government export controls and a security bypass discovered by Amazon researchers. A newly trained safety classifier, the company says, now blocks the specific exploit technique in over 99% of cases.

What's new

Claude Fable 5 became available worldwide on July 1 on the Claude Platform, Claude.ai, Claude Code, and Claude Cowork. Plan details at launch:

• Pro, Max, and Team plans: Fable 5 included for up to 50% of weekly usage limits through July 7, then available via usage credits
• Enterprise: select plans restored on the same timeline
• Cloud platforms: AWS, Google Cloud, and Microsoft Foundry access being re-enabled separately
• Mythos 5: Restored to a set of US organizations following government approval on June 26

The redeployment came less than 24 hours after the US government lifted export controls on June 30.

Context: the bypass and the suspension

Export controls were applied on June 12 after Amazon researchers discovered a method allowing Fable 5 to identify software vulnerabilities and demonstrate exploitation techniques. The concern was specific to Claude Mythos 5's offensive cybersecurity capabilities — Anthropic's own documentation describes Mythos 5 as able to "find and exploit software vulnerabilities more effectively than any other model—and all but the most skilled human security experts."

Because Anthropic could not verify nationality in real time, it suspended access to both Fable 5 and Mythos 5 for all users globally — not just foreign nationals — for the duration of the export restriction.

In response to the Amazon finding, Anthropic trained an improved safety classifier targeting the reported technique. The company reports the classifier blocks the specific bypass in over 99% of cases. Anthropic thanked users "for bearing with us through this disruption."

Why it matters

The Fable 5 suspension is the first known instance of a major US AI lab pulling its flagship model from global access under government export restrictions. The episode establishes a pattern: frontier models with dual-use capabilities — in this case, offensive cybersecurity — are now subject to the same national security controls that govern other sensitive technologies.

For developers and enterprise customers, the practical effect was three weeks of disrupted access to one of the most capable reasoning models available. Cloud platform integrations required separate re-enablement processes, adding operational complexity beyond simply restoring API access.

The security angle raises questions that Anthropic does not fully address: whether a narrow classifier blocking one specific technique constitutes a durable fix, or whether the underlying capability remains accessible through novel prompting approaches. The company's disclosure that Amazon found the issue — and that the fix targets only the reported technique — suggests ongoing adversarial cat-and-mouse dynamics around dual-use AI capabilities.

For the industry, the episode signals that export control risk is now a factor in frontier model deployment planning, and that labs will need clear suspension-and-restore protocols as their models grow more capable.