Pentagon signs classified AI contracts with seven firms, drops Anthropic over supply‑chain risk

The Pentagon moved swiftly on Friday, sealing seven new contracts that permit the use of advanced artificial‑intelligence systems in classified settings. OpenAI, Google, Microsoft, Amazon, Nvidia, Elon Musk’s xAI and the emerging firm Reflection each signed agreements that will embed their AI tools into the Department of Defense’s operational workflow. The announcements signal a concerted push to embed AI across the military’s decision‑making chain, a step the department describes as essential for maintaining a technological edge.

Defense officials stressed that the agreements are limited to "lawful operational use," meaning the AI applications must comply with existing statutes and rules of engagement. By formalizing these partnerships, the Pentagon hopes to streamline access to cutting‑edge models for tasks ranging from data analysis to predictive maintenance, while preserving the security protocols required for handling classified information.

OpenAI and xAI were among the first to reach such deals, and a report from The Information indicates Google has now joined the roster. Microsoft and Amazon, long‑standing contractors, are deepening their existing relationships under the new framework. Nvidia and Reflection represent fresh entries, expanding the portfolio of tools available to analysts and commanders on the ground.

Anthropic excluded amid supply‑chain concerns

Notably absent from the list is Anthropic, a startup that previously secured a $200 million contract to process classified material for the Pentagon. Department officials labeled the company a "supply‑chain risk" after Anthropic refused to ease its red‑line policies that prohibit mass domestic surveillance and fully autonomous weapons. The standoff led Anthropic to sue the government, resulting in a temporary injunction that halted its participation.

Emil Michael, the Defense Department’s chief technology officer, told CNBC that Anthropic’s security model, Mythos, presents a unique national‑security moment. "We have to make sure that our networks are hardened up, because that model has capabilities that are particular to finding cyber vulnerabilities and patching them," Michael said, underscoring the delicate balance between innovative AI capabilities and potential vulnerabilities.

While the Pentagon’s new contracts open a pathway for a broader set of AI tools, officials emphasized that compliance and security remain non‑negotiable. Each vendor will be required to meet strict certification standards before their systems can be deployed in classified environments. The department also indicated that ongoing oversight will monitor how the AI tools are used, ensuring they align with legal and ethical guidelines.

The move comes as the U.S. military seeks to cement its status as an "AI‑first" fighting force, a phrase that has appeared in recent strategic documents. By diversifying its AI supplier base, the Pentagon aims to avoid over‑reliance on any single vendor and to foster competition that could accelerate innovation.

Industry analysts note that the inclusion of both established giants and smaller startups reflects a pragmatic approach to technology acquisition. While firms like Microsoft and Amazon bring mature cloud infrastructures, newcomers such as Reflection may offer specialized capabilities that address niche operational needs.

As the agreements take effect, the Department of Defense will begin integrating the newly cleared AI tools into pilot programs across various branches. The success of these initiatives will likely shape future procurement strategies and could set a precedent for how other federal agencies handle classified AI use.