Anthropic Still Flagged as Supply Chain Risk by Pentagon, US Official Says

The U.S. Department of Defense still lists Anthropic as a supply chain risk to the defense industrial base, according to Emil Michael, under secretary of defense for research and engineering, contradicting Commerce Secretary Howard Lutnick's statement that the AI company is "back on the right side." The conflicting signals follow a federal court ruling on Aug. 27 that found the Pentagon's punishment of Anthropic "illegal and baseless," but the risk label remains, potentially limiting the company's defense contracts.

September 3, 2026, Inside AI — The U.S. Department of Defense still lists Anthropic as a supply chain risk to the defense industrial base, according to Emil Michael, the under secretary of defense for research and engineering. His statement on X came one day after Commerce Secretary Howard Lutnick suggested a thaw in relations. Lutnick told Axios that Anthropic is “back on the right side,” adding, “we trust Anthropic” in an interview. That comment signaled a possible reset between the AI company and the Trump administration. Yet Michael’s post contradicted that impression. He said Anthropic remains flagged as a “Supply Chain Risk” by the Pentagon and defense industry. The Defense Department did not respond to a request for comment. The conflicting signals follow a federal court ruling on Aug. 27. A judge found the Pentagon illegally punished Anthropic for criticizing the government. The ruling called the Pentagon’s decision “illegal and baseless.” Anthropic, founded in 2021, develops large language models including Claude. The company has positioned itself as a safety-focused rival to OpenAI and Google. Its government contracts have grown, but the defense relationship remains strained. The supply chain risk label carries practical consequences. It can limit a company’s access to defense contracts and sensitive programs. For a firm seeking enterprise and public sector growth, that designation matters. The Pentagon’s stance may reflect broader tensions over AI procurement. Defense officials have expressed concern about relying on models they cannot fully audit. Anthropic’s public criticism of certain government policies likely deepened that distrust. Michael’s post did not detail the specific reasons for the continued risk flag. He also did not say whether the label would be reviewed. His statement suggests the Commerce Department’s view does not bind the Defense Department. Lutnick’s comment may have been aimed at reassuring investors and commercial partners. But defense contracting operates under separate rules and risk assessments. A company can be trusted for commerce yet still restricted in national security contexts. The judge’s ruling found procedural and legal flaws in the Pentagon’s punishment. However, a court decision on process does not automatically remove a risk designation. The Pentagon could still maintain the label through a corrected process. Anthropic has not publicly commented on Michael’s post. The company has previously said it wants to work with government agencies on safe AI deployment. It has also emphasized constitutional AI and alignment research. The situation highlights a split within the administration. Commerce sees Anthropic as a partner. Defense sees it as a potential vulnerability. That divide could complicate the company’s federal strategy. For defense contractors, supply chain risk flags can trigger extra scrutiny. They may face limits on data sharing, cloud access, or teaming arrangements. Even a temporary label can affect bidding and reputation. Anthropic’s models are used by some federal civilian agencies. But defense applications require higher assurance levels. The Pentagon often demands on-premise deployment, cleared personnel, and strict audit rights. Michael’s role gives him direct authority over research and engineering policy. His public statement carries weight inside the acquisition community. It signals that the risk flag is not merely a bureaucratic leftover. The timing is notable. Lutnick’s interview appeared designed to project unity. Michael’s post undercut that message within 24 hours. Such public disagreement is rare among senior officials. The defense industrial base includes thousands of companies. Supply chain risk assessments aim to prevent foreign influence, intellectual property theft, and software vulnerabilities. AI models introduce new categories of risk, including prompt injection and data leakage. Anthropic has published research on many of these risks. The company’s safety work is respected in technical circles. But that has not been enough to remove the Pentagon’s flag. Some analysts argue the risk label may be political. The judge’s ruling found the punishment was retaliatory. If so, the label could persist as long as the underlying political friction remains. Others say the Pentagon has legitimate concerns. Large language models can hallucinate, leak training data, or be manipulated. Defense officials may want stronger guarantees before clearing Anthropic for sensitive use. The Commerce Department has no authority over defense procurement risk lists. Lutnick’s trust statement may reflect his own view or White House messaging. It does not bind the under secretary of defense for research and engineering. Anthropic’s next steps are unclear. The company could seek a formal review of the risk designation. It could also work through Congress or the courts. A direct appeal to the secretary of defense is another option. The broader AI industry is watching closely. Other model providers face similar scrutiny. A clear resolution could set a precedent for how defense agencies assess AI vendors. For now, the contradiction stands. One senior official says the company is trusted. Another says it remains a supply chain risk. The result is uncertainty for Anthropic and its potential defense partners.