US Defense Department Partners with Top AI Firms: What This Means for National Security in 2026

[CBC News] The U.S. Department of Defense has signed classified collaboration agreements with seven major AI firms — excluding Anthropic, which restricts military uses of its technology.

On May 1, the Pentagon said in a press release that it reached agreements with SpaceX, OpenAI, Google, NVIDIA, Reflection, Microsoft and Amazon Web Services.

The department plans to field these companies’ advanced AI tools on classified networks for lawful missions. The goal is to accelerate the U.S. military’s transition to an AI‑first force and sharpen decision-making across the battlespace. Pentagon officials also said they will continue building an open architecture to avoid dependence on any single vendor.

The move reflects two realities: Anthropic’s Claude remains the only model effectively integrated into current classified military systems, and Anthropic has clashed with the Pentagon by maintaining bans on large‑scale domestic surveillance and on fully autonomous weapon systems.

Earlier, the Pentagon designated Anthropic as a national security supply‑chain risk. Anthropic responded with a lawsuit, and the dispute is ongoing.

Defense Secretary Pete Hegseth also publicly criticized Anthropic CEO Dario Amodei during a recent Senate Armed Services Committee hearing.

Companies that joined the agreements reportedly consented to allow the Pentagon to use their technologies for all lawful purposes. The New York Times reported the department may also be using the move to pressure Anthropic to alter its policy stance.

Separately, Anthropic has rolled out Mitos, a new model specialized in finding software vulnerabilities, and has continued talks with government agencies. Some agencies are already said to be using the model.

On April 17, CEO Amodei visited the White House and met with Chief of Staff Suzie Wiles and Treasury Secretary Scott Besent. President Donald Trump later described the meeting as “a good meeting.”

But Emil Michael, the Pentagon’s chief technology officer and undersecretary for research and engineering, told CNBC the department still views Anthropic as a supply‑chain risk. He said subordinate units were granted a six‑month transition period and described decisions on Mitos’ use as a separate national security matter for the broader government to resolve.

Michael said the model “specializes in detecting and defending against cyber vulnerabilities,” and added that “we need to use it to strengthen our network defense capabilities.”

★ Vote for the Netizen Awards ▶
★ Breaking news — CBC News Telegram ▶

▮ CBC News | CBCNEWS Reporter Ha Young-soo