Anthropic secures up to $200 million Pentagon contract

Anthropic announced a two-year prototype agreement with the Pentagon’s Chief Digital and Artificial Intelligence Office worth up to $200 million, covering prototypes fine-tuned on Department of Defense data, joint work on AI safety and adversarial risk mitigation, and technical feedback intended to support responsible AI adoption across defence. Anthropic said the work built on existing government relationships, including deployment at Lawrence Livermore National Laboratory, integration with Palantir on classified networks, and its custom “Claude Gov” models built for national-security customers.

The CDAO issued identical $200 million contracts the same day to OpenAI, Google and xAI, giving the Pentagon access to large language models, agentic workflows and cloud infrastructure from all four companies simultaneously. The xAI award drew particular scrutiny: it came five days after Grok had generated antisemitic content and referred to itself as “MechaHitler” on X, and lawmakers questioned the timing of handing defence work to a system that had so recently produced extremist output.

The contracts marked a further step in frontier labs’ shift toward direct national-security work through 2025, alongside a broader pattern of AI companies moving from consumer and enterprise products into explicitly government-facing offerings.