What Happened
President Trump issued an order in late February 2026 directing steps to remove Anthropic from US government procurement, according to Wired reporting. The move followed pressure from the Defense Department on Anthropic to drop restrictions on military use of its AI models - restrictions Anthropic declined to remove. The order came within days of the Pentagon moving toward a supply chain risk designation, escalating to formal executive action faster than most observers anticipated.
The administration's stated position is that Anthropic's use policy restrictions are incompatible with national security requirements and that the company should not receive government business while maintaining them.
Why It Matters
A formal government exclusion from federal procurement would affect Anthropic's business across multiple dimensions. Direct government contracts represent a substantial market. More significantly, the exclusion signals to government contractors - companies that do business with federal agencies and comply with supply chain security requirements - that using Anthropic's products could create procurement complications.
The dispute also tests whether safety-motivated use restrictions are commercially viable when the US government is on the other side. OpenAI and Google have been more accommodating to government and military requirements. This creates direct competitive pressure on Anthropic: hold the policy and absorb commercial consequences, or negotiate restrictions that satisfy the DoD while maintaining credibility with users who trust Anthropic's safety commitments.
For organizations currently using Claude in production, particularly those with federal adjacency, the vendor risk profile has shifted materially within a very short window.
Our Take
This is the most consequential policy story in recent AI news for enterprise buyers. The Anthropic-Pentagon conflict escalated from a policy dispute to executive action in under a week. That pace reflects deliberate political prioritization, not bureaucratic process.
Anthropic's public response - calling the action legally unsound rather than signaling willingness to negotiate - suggests the company views this as a line worth holding. That may prove correct or may prove costly depending on how courts and other agencies respond. For businesses evaluating Claude versus alternatives, this is now a factor in vendor risk assessments that was not present six months ago. Anthropic's API access and pricing have not changed, but the political environment around using the product in government-adjacent contexts has shifted significantly. Organizations with federal contracts should consult legal counsel now on whether vendor selection affects contract eligibility, independent of their own view on the underlying policy dispute.