Pentagon power-user tests deepen as Anthropic litigates exclusion from May 1 contract awards

Bloomberg reports the Pentagon is now testing rival AI models with 25 of the department's 'power users' to identify Anthropic alternatives. The May 1 procurement awards went to OpenAI, Google, Microsoft, AWS, NVIDIA, SpaceX, and startup Reflection AI — Anthropic was excluded after Defense Secretary Hegseth designated the company a supply-chain risk over its refusal of 'all lawful' use language.

The exclusion is the highest-stakes test of safety-first lab posture in the policy regime. Anthropic's argument — that 'all lawful' covers domestic surveillance and lethal autonomous systems — is principled and has substantial bipartisan support among national-security lawyers. The Pentagon's argument — that vendors must accept the same contracting language as defense contractors of comparable scale — is procurement-standard. The court case will resolve which principle wins.

The Q3 2026 implication is structural. If Anthropic loses in court, the safety-first posture becomes a billions-of-dollars revenue cost to maintain — and frontier labs will calibrate their disclose-hold thresholds against that price. The Trump EO signing adds another lever; refusing the framework now carries Pentagon and EO consequences. See our analysis →.

Bloomberg — Pentagon tests rival AI models → · RoboRhythms — Pentagon contracts exclude Anthropic → · Slate — AI Anthropic Claude Pentagon →