The National Security Agency is preparing to deploy Anthropic's Mythos artificial intelligence model in cyber operations, sources told TechCrunch, in an apparent conflict with a federal ban on contracting with the AI company.
The reported move would place one of the most capable commercially developed large language models inside the U.S. government's primary signals intelligence apparatus.
A federal prohibition on using Anthropic as a vendor remains in effect, according to the report published Thursday. No public waiver or executive carve-out has been disclosed.
The NSA did not confirm or deny the preparations. Anthropic declined to comment on any arrangement with the agency.
Intelligence agencies have accelerated AI adoption across signals analysis, code generation, and vulnerability research. Mythos, Anthropic's operational model, has drawn particular interest for its performance on technical and adversarial tasks, sources familiar with the evaluations told the Wail.
Congressional oversight committees have not been publicly briefed on any NSA-Anthropic deployment, according to two people familiar with the matter.
Senate Intelligence Committee members are expected to seek a classified briefing on the reported program as early as next week.