Anthropic's Claude Fable 5 and Mythos 5 get global release after export curbs lifted
US lifts national security restrictions on Anthropic's most advanced AI models after safety testing.
The US Commerce Department has lifted export restrictions on Anthropic's Claude Fable 5 and Mythos 5 models, approximately three weeks after the Trump administration flagged them as national security risks. Commerce Secretary Howard Lutnick notified Anthropic that they 'would no longer need a license for exports or in-country transfers' of these models, acknowledging that Anthropic had 'taken steps in close coordination with the US government to address the risks.' Anthropic had been ordered to shut off global access on June 12 due to fears that China, Russia, or other adversaries could exploit the models for cyberattacks on US infrastructure.
Mythos 5, which possesses 'prodigious cybersecurity capabilities' that can 'find and exploit software vulnerabilities more effectively than any other model,' remains restricted to the Glasswing program for defensive cybersecurity research. Fable 5, designed for the general public, has been redeployed worldwide with enhanced safeguards. Anthropic tightened protections after a jailbreak method (discovered by Amazon researchers) was blocked in 99% of cases, but warned of a 'trade-off' that may cause benign coding prompts to be blocked during routine tasks. The government reserves the right to reimpose curbs if risks re-emerge.
- Anthropic's Claude Fable 5 and Mythos 5 models are now available globally after US lifted export curbs imposed on June 12.
- Mythos 5 is accessible only through the Glasswing program for defensive cybersecurity research due to its unique offensive capabilities.
- Enhanced safeguards on Fable 5 block 99% of jailbreaks but may also prevent some benign coding and debugging tasks.
Why It Matters
Anthropic's release signals a new era of government-industry AI safety cooperation, balancing national security with global deployment.