House Democrat Urges Spy Agencies Get Early AI Access

Nextgov reports that Rep. Jim Himes (D-Conn.), the top Democrat on the House Intelligence Committee, said on a Politico Security Summit panel that it would be "insane" for U.S. intelligence agencies to not have early access to advanced artificial intelligence models. Nextgov reports that the National Security Agency has been testing the model Mythos, which it describes as a major Anthropic model, citing multiple people familiar with the matter. On Monday, The Washington Post reported the Trump administration is split over whether intelligence agencies or the Commerce Department should house an AI evaluation center; Nextgov records Himes saying the Commerce Department should also have a role.

Sensitive, frontier models like Mythos often prompt classified or restricted evaluation pathways because their capabilities intersect with cybersecurity and offense-defense tradeoffs. Industry observers note that when models present dual-use cyber capabilities, agencies with operational responsibility for national defense commonly seek early technical access to assess risks and mitigations.

The debate referenced by The Washington Post reflects an unresolved policy question: which civilian regulator or national-security organisation should set or operationalise evaluation standards for advanced models. For practitioners, the institutional choice affects reporting requirements, export-control interactions, and how vendor-model testing is structured between classified and unclassified environments.

Observers should track any formal proposals or executive actions that define an evaluation center's authority and scope and whether vendors like Anthropic receive structured, nonpublic review pathways. Also watch diplomatic engagements mentioned in the article, including President Donald Trump's China trip, where AI topics are expected to arise.