NSA uses Anthropic Mythos
- Reports say the U.S. National Security Agency is using Anthropic's Mythos Preview despite earlier supply‑chain flags. - Coverage cites experts who value the model's coding and autonomous task capabilities. - The case shows institutions sometimes accept governance contradictions when tool capability materially exceeds alternatives. (economictimes.indiatimes.com)
The National Security Agency is using Anthropic’s Mythos Preview even after the Pentagon labeled the company a supply-chain risk, Axios reported on April 19. (axios.com) Axios said two people with knowledge of the matter described active National Security Agency use of Mythos Preview, Anthropic’s newest high-end model. Reuters matched the broad claim on April 19 but said it could not independently verify the report. (axios.com) (usnews.com) The contradiction is recent and specific. On April 8, a federal appeals court refused to pause the Pentagon’s supply-chain-risk designation against Anthropic while the company’s lawsuit continues. (usnews.com) Reuters reported that Anthropic says Defense Secretary Pete Hegseth applied that label after the company refused to remove guardrails for military uses including U.S. surveillance and autonomous weapons. A California federal judge had blocked a separate order on March 26, but the D.C. case kept the designation in place. (usnews.com) Mythos is not a general public chatbot release. Anthropic introduced it on April 7 as a limited-preview cybersecurity model and said it would keep access narrow because the system’s cyber capabilities could be abused. (cnbc.com) (techcrunch.com) Anthropic’s own product notes describe Mythos Preview as more capable than Claude Opus 4.7 on cyber work, while Opus 4.7 was released more broadly on April 16 with automated blocks for high-risk cybersecurity prompts. Anthropic said it wants to test those safeguards on the less capable model before any wider Mythos-class rollout. (anthropic.com) That helps explain why security agencies want access. CNBC reported on April 17 that Anthropic told officials Mythos is especially strong at finding weaknesses and security flaws in software, and Axios said government officials who dislike the company still see the tools as useful for U.S. cyber operations. (cnbc.com) (axios.com) The White House was already moving in that direction. Reuters and CNBC reported that Anthropic chief executive Dario Amodei met White House Chief of Staff Susie Wiles on April 17 to discuss Mythos, with the White House calling the talks “productive and constructive.” (msn.com) (cnbc.com) Anthropic, the National Security Agency and the Defense Department did not immediately comment in the Reuters pickup of the Axios report. For now, the court fight, the White House talks and the reported National Security Agency use are all moving at once. (usnews.com)
