Notably absent from the agreements is Anthropic. The Pentagon designated the company a supply-chain risk earlier this year after a dispute over the terms under which the military could use its AI tools, barring its use by the Pentagon and its contractors. Speaking to CNBC on Friday, Emil Michael, the Defense Department's CTO, maintained that Anthropic has not been cleared as a supply-chain risk — while singling out the company's Mythos model, which drew alarm for its ability to enhance hacking operations, as what he called a "separate national security moment."
The arrangements build on partnerships that were already underway. Google $GOOGL, for instance, had already been operating within the Pentagon and committed to furnishing the Defense Department with AI agents in an unclassified context earlier this year. More than 600 Google employees signed a letter opposing classified AI work with the Pentagon, citing concerns that such arrangements would prevent workers from knowing how the technology is applied. The signatories called on CEO Sundar Pichai to reject all classified military workloads and create a standing ethics board to evaluate proposed defense deals.
The classified agreements come amid a broader effort by the Pentagon to diversify the range of AI companies working with the military. Senior Defense Department officials have said the military should be permitted to use commercial AI in any situation that is legal under U.S. law and military rules.
The door may still open for Anthropic. Trump offered a more conciliatory signal last week, saying Anthropic was "shaping up," a remark widely read as a possible indication that the administration could lift its ban on the company. Anthropic has sued the administration in California and Washington state seeking to undo the supply-chain designation. A California judge issued a temporary halt to the ban, but the D.C. Circuit Court of Appeals has allowed the designation to remain in place as the litigation continues
