Key facts
- US export controls on Anthropic's Fable 5 and Mythos 5 AI models have been lifted.
- Controls were imposed on June 12 due to cybersecurity concerns after Amazon researchers found a jailbreak vulnerability.
- Anthropic will restore global access to Fable 5 on July 1.
- Mythos 5 will be reintroduced to approved U.S. organizations after government review.
- Anthropic has developed a new safety classifier to block jailbreak techniques, blocking over 99% of attempts.
- Anthropic is collaborating with Amazon, Microsoft, and Google on a jailbreak-severity framework and will provide earlier government access to future models.
The U.S. government has lifted export controls on Anthropic's advanced artificial intelligence (AI) models, Fable 5 and Mythos 5, less than three weeks after they were suspended over national security concerns. Anthropic announced it would begin restoring access to the models on July 1, following notification from the U.S. Department of Commerce that the controls were removed on June 30. The controls were initially imposed on June 12, prompting Anthropic to suspend the platforms globally due to the inability to verify user nationality in real-time. The trigger for the suspension was a cybersecurity finding after Amazon researchers reported a method to bypass Fable 5's safeguards, allowing the model to identify software vulnerabilities and generate exploit code. The administration aims to balance AI innovation with cybersecurity risks, with arrangements tying into a June 2 executive order on AI security. Anthropic is restoring Fable 5 globally, while Mythos 5, a less restricted variant, is being reintroduced to a set of approved U.S. organizations after government review. The company is also deepening its collaboration with the U.S. government on AI model testing and threat intelligence sharing. Anthropic has developed a new safety classifier that blocks the reported bypass technique in over 99% of attempts, with flagged requests rerouted to Claude Opus 4.8. Researchers at the Commerce Department's Center for AI Standards and Innovation reviewed the safeguards. Access details vary by plan, with Pro, Max, Team, and select Enterprise subscribers receiving Fable 5 for a limited period before it shifts to usage credits. Availability on AWS, Google Cloud, and Microsoft Foundry will follow. During the outage, OpenAI's GPT-5.5-Cyber outperformed Mythos 5 on the CyberGym benchmark, and Chinese firms also introduced competing systems. Anthropic is proposing a shared framework for grading future jailbreaks with Amazon, Microsoft, and Google, and has committed to providing the government earlier access to test frontier models.