Pentagon CTO Declares DOD Will Never Again Rely on Single AI Provider Amid Anthropic Dispute

Defense Under Secretary for Research and Engineering Emil Michael stated at the AI+ Expo in Washington on May 7 that the Department of War will "never again be single-threaded with any one model," citing the ongoing Anthropic dispute as the catalyst for a broader vendor diversification push. The department has signed agreements with eight AI developers and is seeking authorization for use across "all lawful use cases"; Oracle Cloud Infrastructure VP Rand Waldron told Defense One the target architecture is four model providers supporting distinct applications from code generation to targeting operations. Anthropic sued the Defense Department and other federal agencies after declining to permit Pentagon use for autonomous weaponry and mass surveillance, and a federal judge ordered the White House to stop directing agencies to remove Anthropic products pending the lawsuit. The U.S. government has separately drafted internal policies permitting agency use of Anthropic's Mythos Preview, which demonstrated the ability to identify cyber vulnerabilities; Michael characterized the capability as "a cyber moment" and treated it as a national security concern distinct from the vendor dispute.