Pentagon Labels Anthropic Supply Chain Risk
The Pentagon has formally notified AI startup Anthropic that it poses a "supply chain risk" to the U.S. after negotiations broke down. Anthropic reportedly demanded "assurances its AI wouldn't be used for mass surveillance or autonomous weapons deployment". Draft U.S. regulations would require companies like Nvidia and AMD to seek government permission for virtually all AI accelerator exports.
Anthropic’s stance is rooted in its "Constitutional AI" approach, a method that bakes a set of safety and ethical principles into its models. These principles are drawn from sources like the UN Declaration of Human Rights and are designed to make the AI "helpful, honest, and harmless" without constant human oversight. The specific "red lines" Anthropic drew with the Pentagon were prohibitions on using its AI for mass surveillance of Americans and for fully autonomous weapons systems that operate without a human in the decision loop. CEO Dario Amodei stated the company would not move on these red lines, arguing the technology is not reliable enough for such applications and could lead to civilian casualties or the deaths of American personnel. In a stark contrast, rival OpenAI secured a deal with the Pentagon for its own models. While OpenAI's CEO Sam Altman has stated they share the same "red lines" as Anthropic, their agreement is structured differently, deferring to existing U.S. law and military policy rather than embedding specific contractual prohibitions that would allow the company to veto certain uses. The "supply chain risk" designation is a legally contentious move, typically reserved for foreign adversaries like Huawei and ZTE. Legal experts argue the designation, based on statute 10 U.S.C. § 3252, is being misapplied as it requires demonstrating a risk of sabotage or subversion by an adversary, not a contractual disagreement with a domestic company. This conflict occurs as Anthropic's valuation has surged. A recent $30 billion funding round valued the company at $380 billion, up from $183 billion just five months prior. The company's annualized revenue has reportedly reached $14 billion, highlighting the significant financial stakes involved in the dispute with the U.S. government. The Pentagon's demand, outlined in a January memo from Defense Secretary Pete Hegseth, is for AI models to be "free from usage policy constraints that may limit lawful military applications." This insistence on an "all lawful use" policy is at the core of the disagreement, with officials stating that military policy should be determined by law, not by the ethical guidelines of a private tech company.
