White House reported nearing Anthropic deal for spy use

Social posts on May 22 and May 23 said the White House was nearing a deal with Anthropic to let U.S. spy agencies use more advanced versions of the company’s AI inside classified systems. The posts named the National Security Agency as a potential user, but no White House announcement, contract filing or agency statement confirming a new agreement was publicly available on Saturday. Anthropic has already said its models are in classified government environments. In a June 6, 2025 post, the company said it introduced “Claude Gov” models built for U.S. national security customers and that those models were “already deployed by agencies at the highest level of U.S. national security,” with access limited to users operating in classified environments. That means the reported talks would appear to be about expanding an existing foothold, not opening an entirely new line of business. Anthropic has not publicly identified the agencies using Claude Gov, and the NSA did not publish a statement on Saturday confirming the social-media claim. ### What has Anthropic already said about classified national-security work? (anthropic.com) Anthropic said on February 26 that it had “worked proactively to deploy” its models to the Department of War and the intelligence community. In the same statement, Chief Executive Dario Amodei said the company was “the first frontier AI company to deploy our models in the US government’s classified networks” and that Claude was being used for “intelligence analysis, modeling and simulation, operational planning, cyber operations, and more.” (anthropic.com) The company’s June 2025 Claude Gov announcement gave more detail on those uses. Anthropic said the models were designed for better handling of classified material, stronger performance on intelligence and defense documents, and improved interpretation of cybersecurity data. ### Where does the NSA fit into this? Axios reported on April 19 that the NSA was using Anthropic’s Mythos Preview model despite a Pentagon designation treating Anthropic as a supply-chain risk, citing two sources. (anthropic.com) Bloomberg reported on April 30 that the NSA had been testing Anthropic’s Mythos model to find cybersecurity vulnerabilities in software including Microsoft products, citing a U.S. official and another person familiar with the matter. (anthropic.com) Those reports do not by themselves confirm a new White House-Anthropic deal, but they do show the NSA has already been linked in published reporting to Anthropic’s more advanced models. Bloomberg also reported on April 16 that the U.S. government was preparing to make a version of Mythos available to major federal agencies, based on a memo it reviewed. (axios.com) ### Why would the White House be involved? The White House has been issuing broader policy guidance to speed federal AI adoption. An Office of Management and Budget memorandum dated April 3, 2025 told agencies to “accelerate the Federal use of AI” while maintaining safeguards for civil rights, civil liberties and privacy. (bloomberg.com) Bloomberg reported on April 30 that White House officials were preparing a national-security AI memo that would set requirements for AI deployment by security agencies. That reporting tied the White House directly to the policy fight over how Anthropic systems could be used in military and intelligence settings. (whitehouse.gov) ### What is the dispute around Anthropic’s government work? Anthropic said on February 27 that negotiations with the Department of War had stalled over two requested exclusions: “mass domestic surveillance of Americans” and “fully autonomous weapons.” The company said it had not received direct communication from the Department of War or the White House on the status of those negotiations at that time. (bloomberg.com) Dario Amodei said a day earlier that Anthropic supported “lawful foreign intelligence and counterintelligence missions” but opposed mass domestic surveillance. That distinction matters because the reported White House talks concern spy-agency use in classified missions, an area Anthropic has publicly described as within scope if the use is lawful and outside the two excluded categories. (anthropic.com) ### What can actually be verified right now? The verifiable facts are narrower than the social posts. Anthropic has publicly confirmed classified deployments, a $200 million ceiling Defense Department agreement announced in July 2025, and extensive work on intelligence, cyber and defense applications. The Office of the Director of National Intelligence also has a standing directive governing how the Intelligence Community develops, acquires and uses AI. (anthropic.com) As of Saturday, the unverified part is the claimed near-term White House deal itself. The next concrete marker would be a White House statement, an Anthropic announcement, or a public filing from a named agency such as the NSA or ODNI. (anthropic.com)