Pentagon signs a classified military network deployment contract with 7 major AI companies: Anthropic still excluded

The U.S. Department of Defense announced on May 1 that it has signed AI deployment contracts with seven technology companies, authorizing their models to run on the most highly classified military networks, Impact Level 6 and Impact Level 7. The signed companies include SpaceX, OpenAI, Google, NVIDIA, Reflection, Microsoft, and Amazon Web Services; just a few hours later, Oracle was also added to the list to become the eighth. Anthropic remains outside the blacklist and did not receive a contract—continuing the standoff between Anthropic’s refusal to loosen safety guardrails for the military.

Contract details: 7+1 companies, Impact Level 6/7 classified networks, 3 major application directions

Impact Level 6 and Level 7 are the highest classified network tiers of the U.S. Department of Defense. The models from the contracted AI companies will be deployed on this tier. There are three target application directions: first, data synthesis, consolidating intelligence from multiple sources quickly; second, enhancing warfighter decision-making; third, improving battlefield situational understanding and awareness. The Department of Defense said that a key design principle of this batch of contracts is “avoiding AI vendor lock” and “ensuring long-term flexibility for the Joint Force,” meaning risks are dispersed by using multiple vendors.

The contracting tech companies cover a wide range of the stack: SpaceX brings in military-grade satellites and computing infrastructure; NVIDIA and AMD (note: AMD is not in this batch, and mainly provides compute capacity through NVIDIA, AWS, Microsoft, and Google) provide the hardware for model training and inference; OpenAI, Google, and Reflection provide cutting-edge models; Microsoft and AWS provide cloud computing and deployment architecture; Oracle (added later) strengthens the enterprise data layer. Taken together, the eight companies already cover the vast majority of core suppliers at the top level of the U.S. AI industry—leaving only Anthropic out.

Anthropic is still excluded: the dispute over military safety guardrails continues

Anthropic’s lack of a contract is a continuation of the Department of Defense’s “supply-chain blacklist” policy toward Anthropic over the past three months. The standoff began in February when Anthropic CEO Dario Amodei refused to allow the military to deploy Claude for autonomous lethal attacks and large-scale domestic surveillance. Defense Secretary Hegseth subsequently categorized Anthropic as a supply-chain risk.

With this 7+1 set of signed vendors, the separation in the industry is made explicit: vendors that “comply with the military’s safety requirements” receive contracts, while vendors that “insist on safety guardrails” are excluded. For Anthropic, although Mythos Preview can still be used by exceptions—such as for NSA—losing the Pentagon’s main contract means losing the scale of government procurement worth billions of dollars over the coming years.

Next to watch: whether Anthropic returns to the negotiation table, and why Reflection was selected at the front of the line

The next thing to watch is whether Anthropic releases concession terms later, seeks entry into the list, and what the Trump administration’s subsequent policy direction will be for Anthropic Mythos through its exception channel. Another focus is Reflection—this comparatively young AI startup being signed alongside OpenAI, Google, and NVIDIA suggests the Department of Defense has assessed its specific capabilities (reportedly, national-security specialization or post–open-source auditability). Its later product disclosures will reflect Pentagon’s specific procurement preferences.

This article Pentagon and 7 AI giants sign a classified military network deployment contract: Anthropic still excluded first appeared on Lian News ABMedia.