Anthropic Deploys Engineers to NSA for Cyber Model Mythos Amid AI Tensions
June 6, 2026
Anthropic has resisted broad use of its models for mass surveillance or autonomous weapons, contributing to a broader mix of government concerns and procurement tensions.
Anthropic has reportedly deployed a small team of forward-deployed engineers to the U.S. National Security Agency to help adapt Mythos, its most capable cyber model, for offensive cyber operations.
The arrangement involves assistance to the NSA in using Mythos, with engineers working to tailor the model for specialized cybersecurity tasks and potential broader use against foreign networks.
This collaboration appears amid tensions between Anthropic and the U.S. government over military use of AI, including a Pentagon procurement ban and regulatory scrutiny cited by DoD designations.
Industry observers note growing competition in AI, with chatter about future OpenAI releases and ongoing debates over balancing capability, speed, and cost.
There is ambiguity and a lack of official confirmation surrounding shifts in government stance toward Anthropic and its technology.
Experts differ on risk and necessity of such capabilities, with some noting that vulnerability-hunting AI has existed under access controls, and others highlighting Anthropic’s confidential IPO plans amid projected rapid revenue growth.
Coverage of the controversy is being reported by major outlets, with FT as a primary source and others providing context.
Affordability remains a central issue, with critics calling Mythos a vanity model due to high costs, while supporters argue it delivers value on tasks needing multiple engineers.
Mythos is described as part of Project Glasswing, which grants extended access to many organizations, with participation growing over time.
Anthropic argues Mythos is not safe for broad release because of capabilities like exploiting zero-day vulnerabilities, leading to restricted access under Glasswing to dozens of vetted groups.
The mid-2020s backstory includes prior Pentagon friction over autonomous weapons and ethical concerns, followed by White House involvement encouraging broader cybersecurity adoption of Mythos.
Summary based on 4 sources
Get a daily email with more Tech stories
Sources

Security Affairs • Jun 6, 2026
Report: Anthropic Deploys Engineers to Support NSA Use of Mythos
Zamin.uz • Jun 6, 2026
Anthropic Launches Secret Claude Mythos Model for US National Security Agency
CXOToday.com • Jun 6, 2026
Anthropic is Now Helping US National Security Agency with Its Mythos Model
WinCentral • Jun 7, 2026
Anthropic’s Mythos AI Sparks Debate Over Unmatched Power and Sky-High Costs