Anthropic Cleared to Supply Claude to the NSA Despite Pentagon Supply Chain Risk Label
Anthropic is set to continue supplying its Claude AI models to the National Security Agency, even after the Pentagon officially labeled the company a supply chain risk. The arrangement has been personally approved by White House Chief of Staff Susie Wiles, according to a New York Times report.
The decision comes down to a hardware problem. US intelligence agencies do not have enough of Nvidia’s latest Grace Blackwell chips to run current AI models from OpenAI and other providers. Anthropic’s newly unveiled “Mythos” model, however, is reportedly capable of running on older chip architecture, making it the only viable short-term option for deployment on the NSA’s classified networks.
The Anthropic Pentagon supply chain risk designation had stemmed from a dispute over how the company’s technology could be used. The Pentagon wanted a broad “any lawful use” clause included in contracts, which Anthropic refused to accept, citing concerns over mass surveillance and autonomous weapons. That language has now been dropped from the deal, with a clause reportedly added to bar the model from processing data belonging to American citizens.
The White House intends to use this contract as a template for future AI agreements with other technology companies. Separately, the administration has also approved $9 billion in new funding to purchase AI chips and resolve the hardware shortage over the long term, though that spending still requires congressional approval.
Anthropic had previously announced a legal challenge to the supply chain risk classification, a move that drew significant attention in the AI and defense sectors. The new deal suggests the two sides have found a workable compromise, at least for now.
