Anthropic Left Out Of Pentagon’s AI Agreements For Classified Work

The U.S. Defense Department has reached agreements with several leading artificial intelligence companies, including OpenAI, Google, Microsoft, Amazon Web Services, Nvidia, SpaceX, and Reflection AI, to support deployment on classified government networks. Nowadays, AI is moving further into the core of state infrastructure, especially in areas tied to defence, intelligence, and sensitive public systems.

One notable absence from the list was Anthropic. Anthropic was left out because of disagreements over usage restrictions and contractual terms. As important as model capability is, this just goes to show that the next phase of AI competition will be shaped by which companies are willing, able, or permitted to work inside high-security state environments.

Defence adoption tends to clarify where institutional trust is actually going. Companies can talk about safety, alignment, and responsible use in public. But when governments begin selecting vendors for classified systems, the criteria become more concrete. Reliability matters. Control matters. Procurement compatibility matters. So does a company’s willingness to operate within state-defined limits. The most important thing in this case is alignment. Antropic didn’t see eye to eye with the Pentagon and their initial deal fell through a few months ago. Their exclusion from these recent agreements comes as no surprise.