Anthropic Restricts Claude Mythos Preview and Defies Pentagon Demands for Unconstrained Military Access
Key Takeaways
- ▸Anthropic restricting Claude Mythos Preview access to Project Glasswing consortium (12+ major infrastructure companies) for defensive vulnerability research
- ▸Claude Mythos has identified thousands of high-severity vulnerabilities in every major operating system and web browser
- ▸Anthropic refusing Pentagon contracts that require unrestricted "any lawful use" of its models; two explicit exceptions: mass surveillance and autonomous weapons
Summary
Anthropic announced it will not broadly release Claude Mythos Preview, its frontier model that has discovered thousands of high-severity vulnerabilities in major operating systems and web browsers. Instead, the company is restricting access to a consortium called Project Glasswing, which includes AWS, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks. The stated goal is defensive: enabling these critical infrastructure companies to find and patch catastrophic flaws before adversaries develop similar capabilities.
The decision represents a notable shift in an industry that has historically prioritized rapid capability deployment. CEO Dario Amodei explained that the company sees frontier models as potentially dangerous to infrastructure, privacy, security, and the public-private boundary. More dramatically, Anthropic publicly announced narrow exceptions to military use of its models: it refuses to support mass domestic surveillance or fully autonomous weapons systems, arguing that AI systems are "not reliable enough" to select and engage targets without human oversight.
The Pentagon responded with pressure and punishment. Refusing to grant Anthropic an "any lawful use" clause in military contracts, Defense Secretary Pete Hegseth designated the company a national security risk, effectively barring it from Pentagon contracts and potentially opening it to broader blacklisting. Anthropic sued for retaliation. As of mid-April 2026, a California federal judge blocked some punitive Pentagon actions and removed stigmatizing labels, while the D.C. Circuit declined to pause a separate supply-chain-risk designation, leaving the litigation unresolved.
- Pentagon retaliated by designating Anthropic a national security risk, blocking contracts and potential blacklisting
- Federal court blocked some Pentagon retaliatory actions, but litigation over supply-chain-risk designation continues in D.C. Circuit
