White House Sidesteps Pentagon Feud to Get Top-Tier AI Into Spy Agencies

The White House has intervened to secure access to advanced artificial intelligence for U.S. intelligence agencies, bypassing a months-long standoff between the Pentagon and AI startup Anthropic. A classified contract now in the works will let the NSA and other spy groups use Anthropic’s latest models, despite the Defense Department labeling the company a supply-chain risk earlier this year.

The move reflects a pragmatic split: intelligence work and military operations demand different rules. The Pentagon wanted unrestricted use of the AI. Anthropic refused, citing concerns about mass surveillance and autonomous weapons. The White House sided with the spies. At the center is a $9 billion secret funding request to buy cutting-edge Nvidia chips and build secure data centers. Without that hardware, agencies risk falling behind adversaries racing to deploy similar capabilities.

Chief of staff Susie Wiles authorized the NSA to keep using Anthropic’s model, called Mythos, while the new contract is finalized. Released in April, Mythos runs efficiently on the latest chips but works on older hardware too—a key advantage when building new data centers takes months. The contract explicitly bars using the AI on Americans’ data, sidestepping the Pentagon’s demand for “any lawful use.”

Intelligence leaders prefer Anthropic’s models for their stronger safety features. The deal could become a template for other AI providers like OpenAI. Chip scarcity complicates everything; Congress still must approve the full $9 billion. Former NSA chief data scientist Vinh Nguyen captured the urgency: “Our intelligence community needs the frontier—the best AI chips, models, systems, talent—on a timeline that matches the threat.”