Anthropic Restores Restricted Access to Mythos 5 Model

Anthropic has resumed limited access to its Claude Mythos 5 artificial intelligence model after a two-week regulatory standoff with U.S. federal authorities. Commerce Secretary Howard Lutnick confirmed that access is now restricted to over 100 pre-approved corporate partners and government agencies. This restoration requires strict adherence to new safety safeguards and federal protocols. The Claude Fable 5 model remains suspended while the company works to meet additional government compliance standards.

The suspension originated from the discovery that the Mythos model could identify vulnerabilities within classified government systems. While officials stated that detection did not lead to an successful exploit, the capability to identify these gaps in hours prompted immediate action. The incident has intensified the global race to develop agentic AI for defense. Reports suggest that China-based researchers are now testing a rival model, GLM-5.2, which reportedly matches the Mythos model’s proficiency in cybersecurity analysis.

This event establishes a global precedent for AI governance. The U.S. is increasingly using executive orders and mandatory audits to control frontier AI. For global policymakers, the challenge is balancing the risks of digital espionage or jailbreaks against the benefits of high-speed vulnerability analysis. India, which is rapidly expanding its public-sector cloud and digital infrastructure, must navigate these emerging security frameworks. Public access to advanced models remains suspended, suggesting a new industry norm where phased, restricted rollouts replace broad launches.