In brief
- Claude Fable 5 is set to return globally on Wednesday, nearly three weeks after a U.S. export control directive forced Anthropic to pull it and sibling model Mythos 5 offline worldwide.
- Anthropic built a new safety classifier that blocks the reported jailbreak technique in over 99% of cases.
- Anthropic is drafting a jailbreak-severity framework with Amazon, Microsoft, and Google, and pledged earlier government access to future frontier models before release.
Claude Fable 5 is coming back. Anthropic said the model will return to users globally on Wednesday across Claude.ai, the Claude Platform, Claude Code, and Claude Cowork, after the U.S. Commerce Department lifted the export controls that had forced it offline since June 12.
Mythos 5, the more capable sibling model built on the same underlying system, will remain limited to vetted partners.
“On Friday, June 12, the U.S. government applied export controls to our newest models, Claude Fable 5 and Claude Mythos 5. This required us to restrict access to foreign nationals, whether inside or outside the United States,” Anthropic wrote in a blog post late Tuesday. “As of today, June 30, the export controls on Fable 5 and Mythos 5 have been lifted.”
The shutdown traced back to a jailbreak report from Amazon researchers, who found a way to bypass Fable 5's guardrails and get the model to identify software vulnerabilities. In one case, the model produced code demonstrating how to exploit one of them. Since the export order applied to all foreign nationals anywhere in the world and Anthropic had no real-time way to verify nationality user by user, the company disabled both models for every user rather than risk noncompliance.
We’ve received notice that the Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5.
We'll begin restoring access tomorrow, and will share an update soon.
We’re grateful to our users for their patience, and to everyone who worked with us on…
— Anthropic (@AnthropicAI) June 30, 2026
In the report, Anthropic explained that its fix was a retrained safety classifier built specifically to catch the reported technique. The company says it now blocks that bypass in more than 99% of attempts, and any flagged Fable 5 request gets automatically rerouted to Claude Opus 4.8 with a notification to the user.
Researchers at the Commerce Department's Center for AI Standards and Innovation reviewed the new and prior safeguards and called them extraordinarily strong, though Anthropic acknowledged the tighter filter will also be prone to false alarms.
In other words, the already strict and overcautious Fable will be even more cautious by design.
Access details vary by plan. Pro, Max, Team, and select Enterprise subscribers get Fable 5 included for up to half their weekly usage limits through July 7, after which it shifts to usage credits, which means the new window is smaller than the original—six days at 50% cap versus 13 days at full inclusion. Ultimately, users are getting less free access than they would have without the ban, even accounting for the second window, saving Anthropic all that computing budget.
Availability on AWS, Google Cloud, and Microsoft Foundry will follow "as quickly as possible," per the company, leaving cloud-hosted enterprise deployments waiting a bit longer than Claude's native apps.
The blackout gave rivals an opening. While Mythos and Fable sat offline, OpenAI's GPT-5.5-Cyber topped Mythos on CyberGym, a UC Berkeley benchmark that tests AI agents against 1,507 known vulnerabilities across 188 open-source projects and scores them by how many they can successfully reproduce.
GPT-5.5-Cyber hit 85.6% against Mythos 5's 83.8%, a gap to which Anthropic couldn't immediately answer since its model wasn't deployable. Chinese firms also used the pause to push their own vulnerability-hunting systems, with Qihoo 360 unveiling Tulong Feng and Z.ai releasing open-weight alternatives.
And smart routing alternatives with weaker LLMs also proved to give similar results to what Mythos reported, even beating it in some cases.
Anthropic is now proposing a shared way to grade future jailbreaks with Amazon, Microsoft, and Google, scoring each on capability gain, breadth, ease of weaponization, and discoverability. It also committed to giving the government earlier access to test frontier models and their safeguards before public release, and opened a new HackerOne program for researchers to report Fable 5 bypass techniques directly.
Daily Debrief Newsletter
Start every day with the top news stories right now, plus original features, a podcast, videos and more.