The National Security Agency is reportedly using Mythos Preview, an advanced artificial intelligence model developed by Anthropic that the company has restricted from public release due to national security concerns. The development raises questions about government access to sensitive technology and the consistency of regulatory positions, particularly as the Pentagon's parent agency simultaneously pursues legal action against the company over alleged security threats.
Anthropicannounced Mythos earlier this month as a frontier model specifically designed for cybersecurity tasks. However, the company determined the model possessed such significant offensive cyber-attack capabilities that public release would pose unacceptable risks. As a result, Anthropic limited access to approximately 40 organizations, publicly naming only a dozen of them. The NSA appears among the undisclosed recipients and is reportedly using Mythos primarily for scanning environments to identify exploitable vulnerabilities.
The Policy Dispute
The NSA's access to Mythos comes amid significant tension between Anthropic and the Department of Defense. The Pentagon labeled Anthropic a "supply-chain risk" after the company refused to grant Pentagon officials unrestricted access to its model's full capabilities. The underlying dispute originated when Anthropic declined to make its Claude model available for mass domestic surveillance and autonomous weapons development—decisions the company characterized as responsible AI governance.
The U.S. military has simultaneously argued in court that Anthropic's tools can threaten national security, creating an apparent contradiction with the NSA's current reliance on the company's most advanced model. The U.K.'s AI Security Institute has also confirmed it has access to Mythos, indicating the model's distribution extends beyond U.S. government agencies.
Shifting Diplomatic Dynamics
Anthropicís relationship with the Trump administration appears to be improving. Three days ago, Anthropic chief executive Dario Amodei met with White House chief of staff Susie Wiles and Secretary of the Treasury Scott Bessent. The White House reportedly characterized the meeting as productive, suggesting potential resolution of earlier tensions between the company and federal agencies.
TechCrunch reached out to the NSA for comment regarding its use of Mythos, but the agency did not respond. Anthropic declined to comment on the matter. The company's decision to restrict Mythos access while simultaneously providing it to the NSA reflects the complex calculus technology firms must navigate when balancing commercial interests, national security requirements, and stated ethical principles.
Why This Matters:
This situation highlights fundamental questions about government consistency in technology regulation and the challenges of managing sensitive AI capabilities. The Pentagon's simultaneous legal action against Anthropic while the NSA actively uses the company's most advanced model suggests unclear policy coordination within the federal government. For the private sector, the episode demonstrates the unpredictability of government technology partnerships—companies face pressure to support national security objectives while the same government institutions may pursue adversarial legal positions. The restricted distribution of Mythos to approximately 40 organizations, with most recipients undisclosed, raises transparency concerns about how government agencies acquire and deploy advanced technology. The thawing relationship between Anthropic and the Trump administration may signal a shift toward more pragmatic technology partnerships, but the underlying tension between national security concerns and commercial AI development remains unresolved.