Pentagon calls Anthropic a national security risk in court filing

The US Department of Defense said Anthropic, an AI lab, poses an “unacceptable risk to national security” in a 40-page filing in a California federal court that rebuts the company’s lawsuits over a Pentagon supply-chain risk label; the filing is the DOD’s first rebuttal to Anthropic’s suits challenging Defense Secretary Pete Hegseth’s decision last month to label the company a supply chain risk.

The filing says Anthropic might attempt to disable its technology or preemptively alter the behavior of its model before or during “warfighting operations” if it deems its corporate “red lines” crossed.

Last summer, Anthropic signed a US$200 million contract with the Pentagon to deploy its technology within classified systems, and in later negotiations over the terms of the contract Anthropic said it did not want its AI systems to be used for mass surveillance of Americans, and that the technology wasn’t ready for use in targeting or firing decisions of lethal weapons.

The Pentagon said a private company should not dictate how the military uses technology while several tech companies and employees – including from OpenAI, Google, and Microsoft – as well as legal rights groups filed amicus briefs supporting Anthropic.

Anthropic’s lawsuits accuse the DOD of violating its First Amendment rights and of acting on ideological grounds, and seek a preliminary injunction, with a hearing set on March 24.

🔗 Source: TechCrunch

🧠 Food for thought

Implications, context, and why it matters.

The Pentagon’s risk label is a targeted procurement tool, not a total ban

Public remarks suggested a sweeping ban. The “supply chain risk” label is a procurement authority that lets the Pentagon limit purchases, not a wider sanction 1.
Under the law, it blocks Anthropic products, including Claude, only in certain “National Security Systems” (classified or mission-critical government systems) used for intelligence, command and control, and weapons systems. It does not cover routine administrative or business applications 1.
Anthropic plus partners such as Microsoft, Google, and Amazon say the label still allows contractors to use Claude for commercial work outside tasks tied to their Department of Defense contract performance 2.
Earlier supply-chain-risk labels aimed to keep foreign technologies out of sensitive US military systems to preserve operational security (the ability to keep military operations secure). This time, the government applies the tool to a US company during a contract dispute over terms, including “all lawful use” provisions 3 1.

This dispute could chill AI safety innovation across the industry

The designation may shift leverage between Silicon Valley and Washington, leaving tech firms with less room to negotiate contract terms 4.
Experts and industry groups warn the move could cool innovation. Companies may hold back on safety guardrails if that invites exclusion from government markets 4 3.
A public “scarlet letter” as a supply chain risk could hurt Anthropic’s commercial business, since some non-government clients may worry about the association 4.
The standoff reopens a debate over who carries responsibility for AI safety. Anthropic cites ethical red lines while the Pentagon seeks final say over all “lawful uses” of technology it buys 5.