Pentagon CTO Emil Michael reinforced that Anthropic is still barred from Department of Defense systems amid ongoing cybersecurity evaluations of Mythos by some federal agencies. Despite speculation about a warming relationship following recent developments, Anthropic remains off-limits given supply chain risk considerations.
- Anthropic still considered a national security supply chain risk
- Mythos AI model evaluated but not deployed by some federal agencies
- Government aims to understand and mitigate AI cybersecurity risks
What happened
Pentagon Chief Technology Officer Emil Michael publicly stated that Anthropic continues to be regarded as a supply chain threat and is not permitted within Department of Defense systems. This clarification came amid speculation that recent activity suggested a relaxation in the government’s stance, particularly following the National Security Agency’s reported use of Anthropic’s Mythos model for evaluation.
Michael emphasized that any use of Mythos at federal agencies remains strictly for assessment purposes rather than operational deployment. He also highlighted that other frontier AI models, including those from foreign sources, are being examined under similar scrutiny to understand their capabilities and cyber risk implications.
Why it matters
The firm position on Anthropic signals continued caution around integrating AI technologies considered potential cybersecurity liabilities within critical federal systems. Supply chain risks in AI could expose networks to vulnerabilities if untrusted components or models are introduced without adequate security assurance.
Moreover, the evaluation of Mythos reflects a broader government effort to stay ahead of emerging AI threats by analyzing powerful models capable of identifying cyber weaknesses. This approach aims to bolster federal network defenses by preemptively understanding novel AI capabilities, while avoiding premature adoption of unvetted technologies.
What to watch next
The federal government is expected to continue examining AI models including Mythos and other newly introduced frontier models, such as ChatGPT 5.5-Cyber, with an eye toward their cybersecurity applications and risks. Upcoming meetings between AI industry leaders and White House officials will likely focus on developing frameworks to securely integrate AI innovations across public and private sectors.
Observers should monitor any changes in policy regarding Anthropic and similar vendors, especially if risk assessments evolve. Additionally, how the Pentagon and other agencies balance AI evaluation with operational deployment will be critical in shaping future AI adoption strategies amidst ongoing national security concerns.