IC Oversight & Authorities — 2026-05-13
US Intelligence Agencies Push for Greater Role in AI Regulation as White House Prepares Executive Order
Commerce Department officials are resisting a White House proposal to house an AI model evaluation center within the intelligence community, the Washington Post reported May 11. The regulatory contest centers on whether the IC's classified assessment capabilities or Commerce's existing standards infrastructure should lead evaluations of frontier AI models for national security risks. Federal News Network reported May 12 that the White House is actively studying an AI security executive order, and Defense One reported that intelligence agencies have already secured informal testing access to frontier models, establishing operational precedent the executive order would formalize or constrain.
Analysis
The interagency contest over AI model evaluation authority reflects a deeper disagreement about whether national security imperatives or market-regulation norms should govern frontier AI access. The NSA's testing ofMythos , an Anthropic model withheld from public release due to its cyber capabilities, shows the intelligence community already exercises informal access to the most sensitive frontier models. The regulatory fight is about formalizing that access on IC terms. Primary evaluation authority for the IC is unlikely to materialize in the executive order within the next 90 days, when the executive order is expected to be issued. Commerce is actively resisting the proposal, the draft order reportedly omits mandatory testing requirements that would anchor such authority, and even the IC's congressional advocates favor shared oversight over IC primacy. Confidence in this assessment is high, grounded in convergent independent reporting from multiple outlets covering distinct nodes of the policy debate. The Anthropic-DOD legal dispute adds a complicating variable: prolonged friction between Defense and the country's most capable AI developer could erode the informal access arrangements the IC currently relies on, making formal regulatory authority both more urgent and harder to secure.
The interagency contest over AI model evaluation authority reflects a deeper disagreement about whether national security imperatives or market-regulation norms should govern frontier AI access. The NSA's testing of
3 sources
- In Trump administration battle over AI US spy agencies seek more power -
Washington Post - 'It would be insane' for spy agencies to lack early access to AI models, lawmaker says -
Defense One - WH 'studying' AI security executive order -
Federal News Network