The US administration has significantly expanded its network of artificial intelligence (AI) suppliers for classified operations, with the Pentagon recently signing agreements with tech giants Microsoft, Amazon, and Nvidia, alongside Reflection AI, which has yet to release a publicly available model. These companies now join established partners like OpenAI, xAI, and Google, enabling the Department of Defense to deploy their AI products for “any lawful use.” This strategic expansion is designed to enhance national security capabilities and foster an “AI-first fighting force” by providing advanced tools to warfighters.

Central to this discussion is the contentious relationship with Anthropic AI, whose CEO, Darius Amodei, raised strong objections to the “any lawful use” clause. Amodei expressed concerns that it could facilitate the US government’s use of Anthropic technology for surveillance of American civilians and the development of autonomous weapons, areas he explicitly sought to restrict. This disagreement led to the Pentagon canceling a $200 million contract with Anthropic, prompting the company to sue for millions in lost revenues. The Trump administration subsequently labeled Anthropic a “supply chain risk” – an unprecedented status for a US-based company – with government sources further describing it as “woke.”

Despite the controversy, the Pentagon’s rationale for broadening its AI vendor base is clear: to prevent “AI vendor lock-in” and ensure “long-term flexibility for the Joint force.” The new technologies are intended to give warfighters the tools they need to act with confidence and safeguard the nation against various threats. These AIs are earmarked for 'Impact Levels' six (secret data) and seven (the most highly-classified materials), moving beyond the current scope of generative AI use, which is largely confined to non-classified tasks like document drafting, summary, and research within defense departments. The expanded capabilities will enable defense forces to “streamline data synthesis,” “elevate situational understanding,” and “augment warfighter decision-making in complex operational environments,” though it remains unclear if these applications extend to domestic deployments within US borders.

This expansion also serves a crucial strategic purpose: to make the US military and security forces more resilient to potential “apparent changes of heart by individual vendors” that could otherwise impact critical operations. By diversifying its technological foundation, the Pentagon aims to mitigate risks associated with the “personal whims of individual company leaders.” This concern is not unfounded, as both Google and Amazon have previously faced internal protests and even fired employees over their companies’ technology being used in weaponry and warfare contexts.

Intriguingly, despite the public fallout, Anthropic AI’s technologies retain a significant presence in US government operations. While new suppliers may replace Anthropic’s Claude AI’s past role in handling classified material via Palantir’s Maven toolset, Anthropic’s Mythos model is reportedly still utilized by the National Security Agency (NSA) for cyber warfare and defense capabilities. Worldwide, Mythos is under assessment by 40 organizations, with the UK’s MI5 and the US NSA believed to be among them. Recent reports from Axios suggest a potential thawing of relations, with sources indicating the US administration is seeking ways to “save face and bring ’em back in.” Furthermore, Anthropic’s Claude coding model has allegedly remained in continuous use by US government security organizations throughout these events. The White House has reiterated its commitment to “proactively engage across government and industry to protect our country and the American people, including by working with frontier AI labs,” underscoring the complex and evolving landscape of AI integration in national defense.