On June 6, 2025, Anthropic announced Claude Gov, a custom set of Claude models built specifically for US national-security customers. Access is limited to those who operate in classified environments, and Anthropic said the models were “already deployed by agencies at the highest level of U.S. national security.”
The company described several ways the Gov models differ from its commercial offerings. They feature “improved handling of classified materials, as the models refuse less when engaging with classified information,” “greater understanding of documents and information within the intelligence and defense contexts,” “enhanced proficiency in languages and dialects critical to national security operations,” and “improved understanding and interpretation of complex cybersecurity data.” Anthropic said the models were shaped by direct feedback from government customers and went through the same safety testing as its other systems.
The “refuse less” framing drew particular attention. Anthropic’s consumer models are tuned to decline many requests involving sensitive or potentially harmful content; the Gov models are deliberately adjusted to be more cooperative with classified material that government users are authorized to handle. The launch put Anthropic in direct competition with other labs offering government-specific systems, and it formed part of a broader push that also included a Pentagon prototype agreement and expanded access for government customers.
For a general reader, Claude Gov shows how AI providers tailor the same underlying technology to very different customers, including loosening guardrails for users operating under their own legal authorities. It is a concrete example of how “responsible AI” gets redefined as a product moves from the public into the national-security domain.