"These agreements accelerate the transformation toward establishing the United States military as an AI-first fighting force and will strengthen our warfighters’ ability to maintain decision superiority across all domains of warfare," the Pentagon said in an initial announcement on the deals announced May 1, 2026.
Which companies are approved to deploy on classified networks
The Pentagon announced agreements with eight technology firms to deploy their AI systems on classified networks. An initial press release listed seven firms: Amazon Web Services, Google, Microsoft, OpenAI, SpaceX, NVIDIA and Reflection — a newer startup backed by NVIDIA. Later the Pentagon CTO’s office posted on X that "Oracle has officially agreed to join the list of AI companies deploying frontier capabilities on the Department’s classified networks."
What classification levels were cleared: Impact Level 6 and Impact Level 7
The announced approvals cover networks designated Impact Level 6 (IL6) and Impact Level 7 (IL7). The release defines IL6 as networks handling secret data and describes Impact Level 7 as "a semi-official term for the most highly classified systems."
Pentagon rationale and officials' remarks
The announcement framed the agreements as a way to "streamline data synthesis, elevate situational understanding, and augment warfighter decision-making in complex operational environments." The move builds on earlier Defense Department activity: the standup of the secure but unclassified GenAI.mil platform in December, and an explicit push by Secretary Pete Hegseth to bring commercial AI into the department.
Pentagon CTO Emil Michael, described in the release as "the under secretary for research & engineering," discussed the initiative on CNBC and cast the effort as one of supply diversification. Michael said, in part: "What we’ve learned since we started this effort at the Department of War is that it’s irresponsible to be reliant on any one partner, and we learned that that one partner didn’t really want to work with us in the way we wanted to work with them." He added the department secured providers across open-source, proprietary-model, and infrastructure categories to "make sure we had diversity of supply."
Anthropic's absence and the National Security Agency note
Notably absent from the Pentagon's announced list is Anthropic. The story states that Anthropic's Claude AI was already in use on classified networks as part of Palantir’s Maven toolkit, but that "the administration has tried to ban [Anthropic] from government work, leading to a brace of lawsuits." Despite that administrative effort, the report says the National Security Agency is "reportedly using Anthropic’s new and not yet publicly available Mythos model," which the source characterizes as "said to have significant cyber warfare capabilities."
Availability timeline and payment terms remain unspecified
The Pentagon announcements did not specify when the newly approved models would go live on IL6 or IL7 networks, nor did they disclose whether the companies would be paid and, if so, how much. The initial release emphasized operational benefits for "warfighters’ ability to maintain decision superiority," but left timing and commercial terms unaddressed.
What this means for warfighters, Pentagon procurement, and the NSA
- Warfighters: The Pentagon frames the approvals as tools to "elevate situational understanding" and augment decision-making on classified networks, signaling intended operational integration of commercial frontier AI into mission workflows.
- Pentagon procurement and engineering teams: The department says it sought a diversity of supply across open-source and proprietary providers and infrastructure companies, reflecting a procurement posture aimed at reducing dependence on any single supplier.
- National Security Agency: The NSA is named in the reporting as a user of Anthropic's Mythos model despite an administration-level ban on Anthropic for government work, highlighting parallel and possibly conflicting acquisition uses within the national security enterprise.
The Pentagon has broadened its set of commercial AI partners for classified networks to eight firms and framed the step as central to making the military "AI-first." The announcement answers which vendors are to be allowed on IL6 and IL7 systems, but leaves open when those models will be operational and whether commercial terms exist. It also highlights a classified-environment split: Anthropic is publicly excluded from this set even as the NSA is reported to use Anthropic’s not-yet-public Mythos model — a contrast the Pentagon's CTO himself acknowledged by stressing supplier diversity.
