Deutsch
English
Español
Français
Nederlands
Português
Anthropic Restores Claude Fable 5 Globally After US Government Lifts Export Controls
On June 30, 2026, Anthropic announced the redeployment of its advanced AI model, Claude Fable 5, after the US government lifted export controls that had been imposed on June 12. The restrictions, which affected both Fable 5 and the more powerful Mythos 5, were triggered by a report from Amazon researchers detailing a method to bypass Fable 5's safety safeguards. The 2026 AI export controls required Anthropic to restrict access to foreign nationals immediately, leading to a global suspension of both models.
Background: What Led to the Export Controls?
On Friday, June 12, the US government applied export controls to Claude Fable 5 and Claude Mythos 5 under a new national security framework. The order came after Amazon researchers discovered a technique that could prompt Fable 5 to identify software vulnerabilities and, in one case, produce code demonstrating how a vulnerability could be exploited. Although Anthropic's testing confirmed that many less capable models—including Claude Opus 4.8, GPT-5.5, and Kimi K2.7—could produce similar results, the government acted swiftly to prevent potential misuse.
Anthropic responded by suspending access to both models for all users, as the immediate effective date of the order left no time to implement nationality verification systems. The company then worked closely with the US Department of Commerce's Center for AI Standards and Innovation (CAISI) to develop improved safeguards.
Timeline of Events and Safeguard Updates
Anthropic released Fable 5 and Mythos 5 on June 9, 2026. Fable 5 was designed with strong safeguards for general use, while Mythos 5—with fewer restrictions—was reserved for trusted Project Glasswing partners focused on defensive cybersecurity. The export control directive on June 12 followed the Amazon researchers' findings.
Over the following two weeks, Anthropic trained an improved safety classifier that targets and blocks the specific behavior described in the Amazon report. The new classifier now blocks the reported bypass technique in over 99% of cases. Researchers from CAISI have tested both the prior and new safeguards and confirmed they are "extraordinarily strong". However, the new classifier also increases false positives during routine coding tasks, which Anthropic says it will continue to refine.
How the Safety Classifiers Work
Anthropic uses a defense-in-depth approach with multiple safety mechanisms. One key component is classifiers—smaller automated AI systems that detect when a model is asked to perform potentially harmful cybersecurity tasks. These classifiers block the model from responding to dangerous requests. The company deliberately sets a "safety margin," blocking some benign requests to ensure that genuinely harmful ones are caught. For Fable 5, this margin was made much larger than in any prior launch, accepting more false positives in exchange for stronger safety.
A New Industry Framework for Assessing AI Jailbreaks
The incident highlighted the lack of a consensus standard for evaluating AI jailbreaks—techniques that bypass a model's safeguards. Anthropic, in partnership with Amazon, Microsoft, Google, and other Glasswing partners, has begun developing a shared framework to score jailbreak severity on four criteria:
- Capability gain: How far beyond existing tools does the jailbreak take the user?
- Breadth of capability gain: For how many distinct offensive tasks does the same technique work?
- Ease of weaponization: How much human effort is needed to turn the jailbreak into an attack?
- Discoverability: How easy is it for someone to obtain the technique?
This framework aims to help AI developers triage findings, prioritize fixes, and communicate risks consistently to governments and industry partners. Anthropic has also launched a new HackerOne program for security researchers to submit potential cyber jailbreaks. The AI safety industry standards initiative is expected to evolve with feedback from additional partners.
Deeper Government Collaboration on Frontier AI Security
Anthropic is strengthening its collaboration with the US government, building on nearly two years of pre-existing partnerships. The company has committed to four key areas:
- Pre-release government access and evaluation: Providing designated partners with early access to models and safeguards for independent testing.
- Rapid information sharing: Notifying government counterparts about significant jailbreaks or misuse patterns and sharing new safeguards.
- Dedicated resources for joint research: Scaling up joint work on AI security with dedicated teams and compute allocations.
- A common industry bar: Working toward a shared voluntary security and evaluation standard for frontier model providers.
Anthropic stated that these rules should be "codified in strong regulation and applied equally across frontier model developers" to provide certainty about access to powerful models. The US AI regulation 2026 developments reflect a broader push for systematic oversight of advanced AI capabilities.
What's Next for Fable 5 and Mythos 5
Fable 5 became available globally on July 1, 2026, on the Claude Platform, Claude.ai, Claude Code, and Claude Cowork. For Pro, Max, Team, and select Enterprise plans, Fable 5 is included for up to 50% of weekly usage limits through July 7, after which it will be available via usage credits. Access on AWS, Google Cloud, and Microsoft Foundry is being restored as quickly as possible.
Mythos 5 access has been restored for a set of US organizations following government approval on June 26. Anthropic continues to coordinate with the government to expand access to broader domestic and international partners in the Glasswing program.
Frequently Asked Questions
What was the reason for the US export controls on Claude Fable 5?
The controls were applied after Amazon researchers found a method to bypass Fable 5's safeguards, prompting it to identify software vulnerabilities and produce exploit code. The government acted under a new national security framework to prevent potential misuse of the model's advanced cybersecurity capabilities.
Are Claude Fable 5 and Mythos 5 available again?
Yes. Fable 5 became available globally on July 1, 2026. Mythos 5 access has been restored for approved US organizations, with plans to expand to international partners in the Glasswing program.
What safeguards were added to Fable 5?
Anthropic trained a new safety classifier that blocks the reported bypass technique in over 99% of cases. The company also expanded its safety margin, blocking more benign requests to ensure harmful ones are caught, and continues to refine its classifiers.
What is the new industry framework for jailbreaks?
Anthropic, Amazon, Microsoft, Google, and other partners are developing a consensus framework that scores jailbreaks on capability gain, breadth, ease of weaponization, and discoverability. This will help standardize responses to AI safety findings.
How will Anthropic collaborate with the US government going forward?
Anthropic has committed to pre-release government access, rapid information sharing on safeguards, dedicated joint research resources, and working toward a common industry security standard for frontier AI models.
Follow Discussion