Claude Gov

Claude Gov is Anthropic’s family of large language models fielded for United States national security work, running inside classified environments where the commercial Claude service is not permitted to reach. Anthropic announced the line publicly in June 2025, describing it as a set of custom models built for a narrow customer base — agencies already operating at the top-secret tier — after roughly a year of quieter deployment through partners.

The technical positioning is straightforward: the same underlying model family as the commercial Claude releases, retrained and guardrail-tuned for defence and intelligence workflows. Anthropic states that the Gov variants show improved handling of classified material, greater fluency in the languages and dialects central to national-security work, and refined understanding of intelligence documents. The models also refuse fewer prompts touching on classified topics — a behavioural shift the company describes as necessary for the work while keeping standard safety training in place elsewhere. No data flows back to Anthropic from a Gov deployment; the model sits inside the customer environment.

Distribution runs through two commercial routes. The first is Palantir — Anthropic’s models were integrated into Palantir’s AIP platform in November 2024, giving users on the DoD’s Impact Level 6 (IL6) cloud access to Claude through Palantir’s tooling. The second is AWS GovCloud and, via Amazon Bedrock, the wider set of United States government cloud regions carrying FedRAMP High accreditation. Both channels predate the branded Claude Gov release and continue alongside it.

Publicly named customers are limited. Palantir’s own announcements point to Department of Defense users, and press coverage has cited unnamed intelligence agencies as early adopters, though specific programme names have not been disclosed. Anthropic has said the models are already in operational use at the highest levels of United States national security. The company’s broader defence posture — a Usage Policy update in June 2025 permitting work with military and intelligence customers, plus Palantir contracts adjacent to Project Maven — sits behind the product.

Claude Gov puts Anthropic alongside OpenAI, whose ChatGPT Gov launched in January 2025, and long-established defence AI vendors in a market whose commercial contours are still forming. The distinctive element is the delivery stack: rather than sell direct, Anthropic reaches the classified customer through Palantir’s data platform and Amazon’s compliant cloud, letting each partner supply the accreditation and integration work while the model itself carries Anthropic’s name.