U.S. Federal Court Upholds Pentagon's Exclusion of Anthropic from Military Procurement

The U.S. Federal Court of Appeals upheld the Pentagon's decision to exclude AI startup Anthropic from military procurement. The ruling legally validated the "cybersecurity supply chain risk" designation that the Department of Defense had assigned to Anthropic.

The issue originated when the Department of Defense excluded Anthropic's AI technology from military contracts. Supply chain risk refers to the assessment that procured products or services could pose a security threat, and it is a legal framework used by government agencies to exclude specific companies from contracts. Defense Secretary Hegseth argues that the safety restrictions Anthropic has implemented in its products—such as the functionality to refuse generating harmful content—could impede military operations.

Meanwhile, Anthropic has strongly opposed this designation. The company claims that this measure has already caused damages in the tens of billions of dollars, and argues that its exclusion from military and government-related contracts has resulted in a significant loss of business opportunities. The appellate court did not accept its arguments in this review, and the Pentagon's decision was upheld.

Anthropic is widely recognized as an AI company in competition with OpenAI, known for ChatGPT, and has championed a corporate policy of "prioritizing safety." Ironically, this safety policy has been identified as a barrier to military contracts, highlighting a difficult question about how corporate ethical design decisions are evaluated in government procurement.

This issue demonstrates a fundamental structural tension in the relationship between AI companies and military and government agencies. While government and military actors seek "AI that follows directives without restrictions," most commercial AI companies intentionally incorporate mechanisms to limit use cases, either to avoid social criticism or for ethical reasons. How to reconcile these conflicting demands through legal and institutional frameworks is positioned as a challenge for the entire AI industry.

Going forward, attention will focus on whether Anthropic will consider appealing to a higher court and whether this ruling will have implications for how other AI companies conduct business with government and military entities. The question of how AI companies can balance "safety considerations" with "government collaboration" remains an issue the entire industry must grapple with, and debate on this matter is expected to continue.