The Pentagon’s decision to label Anthropic’s Claude a supply chain risk signals a major escalation in the fight over military AI, with the core issue now less about model capability than who controls the values and constraints embedded inside the system. (Source: Image by RR)

Defense Contractors Will Be Required to Certify That They Do Not Use Claude

The Pentagon’s top technology official has offered its clearest public justification yet for blacklisting Anthropic, saying the company’s Claude models would “pollute” the defense supply chain because their underlying values and policy constraints are built directly into the system. Defense Department CTO Emil Michael argued that the military cannot rely on models whose internal guardrails reflect what he described as a different set of policy preferences, particularly when those systems may influence mission-critical defense applications.

The designation, as noted in cnbc.com, is extraordinary because Anthropic is the first American company to be publicly labeled a supply chain risk, a category more commonly associated with foreign adversaries. Under the new rules, defense contractors and vendors will need to certify that they are not using Claude in Pentagon-related work. Anthropic has responded by suing the Trump administration, calling the move unlawful and warning that hundreds of millions of dollars in contracts are now in jeopardy.

Michael’s remarks suggest the core dispute is not simply technical performance but governance and control. Anthropic’s models are shaped by a published “constitution” designed to guide behavior around safety, ethics, honesty, and sensitive information. The Pentagon appears to view those embedded principles as a strategic liability if they could constrain military decision-making or system outputs in ways the Department of Defense did not choose. Anthropic, meanwhile, argues that its safeguards are part of responsible AI deployment, not a national security threat.

Despite the blacklist, the transition away from Claude will not be immediate. Pentagon officials acknowledge Anthropic’s technology is already embedded in defense workflows and cannot simply be removed overnight. Palantir CEO Alex Karp said his company is still using Claude, and Michael himself conceded the Defense Department will need time to shift to another vendor. The standoff now sits at the center of a larger battle over who gets to define acceptable AI behavior in military systems: the model maker, the contractor, or the state.

read more at cnbc.com