"We were single-threaded on one vendor, one AI vendor at the Department of War," Emil Michael said, "and to integrate into classified systems is not just putting your software on a public cloud and having it work." The defense undersecretary for research and engineering used that line Thursday to underscore a new procurement posture: the Pentagon will not again depend on a single artificial-intelligence provider.
Emil Michael framed the shift at the AI+ Expo
Speaking at the Special Competitive Studies Project’s AI+ Expo in Washington, D.C., Emil Michael described the Pentagon’s reliance on a single vendor as a hard lesson. He said recent agreements with eight leading AI developers are a "counterstatement" to the ongoing conflict with Anthropic and a move toward more flexible contracts. Michael emphasized the technical difficulty of integrating AI into classified systems and repeated the commitment: "never again we'll be single-threaded with any one model."
Anthropic’s legal fight and the rise of Mythos Preview
The Pentagon’s decision to diversify vendors is occurring against the backdrop of a legal dispute with Anthropic. Anthropic has sued the Defense Department and other federal agencies and their leaders after being declared a national-security risk when it declined Pentagon requests to let its tools be used for autonomous weaponry and mass surveillance of Americans. A judge has ordered the White House to stop telling agencies to remove Anthropic's products while the lawsuit proceeds.
At the same time, Anthropic released a new product, Mythos Preview, that the Defense Department has taken an interest in for its cyber capabilities. Michael called that development "a cyber moment," and asked rhetorically, "How is the U.S. government going to deal with cyber?" The U.S. government has drafted internal policies that would allow agencies to use Mythos, according to the account Michael gave.
Pentagon agreements with eight AI developers and the promise of "all lawful use cases"
Michael described the Pentagon’s recent deals with eight leading AI developers as both a sign of industry support and a contractual promise. He said the agreements represent a willingness by companies that had previously shied away from military work to let the Pentagon use their tools "for all lawful use cases." The administration of multiple, distinct models is intended to prevent the integration bottleneck that followed the earlier single-provider approach.
Oracle’s Rand Waldron warns on interconnectedness and use-case specialization
Rand Waldron, vice president of the Global Government Sector for Oracle Cloud Infrastructure, told Nextgov/FCW that Defense officials are asking cloud service providers like Oracle to prioritize interconnectedness to avoid future vendor lock-in. Waldron said the Pentagon expects that different models will be better suited for distinct missions—examples he offered include code generation, data analytics, supply chain management or targeting in warfighter operations—and that "all those different use cases will [not] end up being the exact same model at any given time."
Waldron noted precedent for the Pentagon’s flexible approach, citing the creation of the Commercial Cloud Enterprise and the Joint Warfighting Cloud Capability contracting vehicles as blueprints for future AI contracting structure. He summarized the Pentagon’s intent plainly: "It's not like they're trying to replace Anthropic with another model provider... They want to replace Anthropic with four model providers."
What this means for technologists, policymakers, and cloud providers
- Technologists and security teams: They will face the operational task of integrating multiple, specialized models into classified systems—an effort Michael said is technically demanding and not equivalent to "putting your software on a public cloud and having it work."
- Policymakers and procurement leaders: They now have a stated mandate to design flexible contracts and acquisition vehicles that permit multiple providers to serve different use cases, drawing on the Commercial Cloud Enterprise and Joint Warfighting Cloud Capability as contracting precedents.
- Cloud providers and AI vendors: Firms such as Oracle are being pressed to build interconnected offerings that reduce lock-in, and vendors previously reluctant to engage in defense work appear to have signaled willingness to permit "all lawful use cases" under the new agreements Michael described.
Conclusion
The Pentagon’s announced shift is explicit and pragmatic: avoid single-provider dependence, contract multiple specialty models, and insist on interoperability. That posture responds directly to the Anthropic dispute and to the Pentagon’s interest in tools such as Mythos Preview for cyber operations. The choice now facing defense acquisition—how to stitch together multiple providers into a secure, serviceable whole for classified missions—was framed by Emil Michael as both a technical challenge and a policy imperative that "never again" will the department be single-threaded on one AI model.
