Pentagon Official Says Anthropic Remains Flagged as Supply Chain Risk, Despite Commerce Secretary's Praise

When a defense official and a cabinet secretary contradict each other publicly within 24 hours, it's usually a sign that something bigger is unsettled behind the scenes. That's exactly what happened this week with Anthropic, the AI company behind the Claude family of models, as two senior figures in the Trump administration offered starkly different accounts of where the firm stands with the U.S. government.

Emil Michael, the under secretary of defense for research and engineering, posted on X that Anthropic is still classified as a "Supply Chain Risk" by the Defense Department and the broader defense industrial base. That designation matters. It signals to contractors, suppliers, and military planners that a company's products or partnerships carry enough uncertainty or potential vulnerability that extra caution is warranted, whether that's about data security, foreign influence, or reliability in a crisis.

Michael's post came just a day after Commerce Secretary Howard Lutnick told Axios that Anthropic is "back on the right side," adding that "we trust Anthropic." Those are not small words from a top economic official whose department oversees export controls and much of the government's approach to AI competitiveness. For a company that builds some of the most capable AI systems in the world, being labeled trustworthy by one senior official and risky by another, within the same week, tells you the government itself hasn't settled the question.

A Legal Fight That Set the Stage

The dispute didn't come out of nowhere. On August 27, a federal judge ruled in Anthropic's favor, determining that the Pentagon had illegally punished the company for criticizing the government. The judge's language was blunt: the Pentagon's decision was "illegal and baseless." That's a serious rebuke. Courts don't often use words like "baseless" to describe federal agency conduct without strong evidence behind it.

Think of the Supply Chain Risk designation like a credit flag on a company's file. Once it's applied, it can follow a firm through procurement decisions, contract reviews, and partnership vetting, even after the original justification is challenged or overturned. A judge striking down the reasoning behind a designation doesn't automatically erase the designation itself. That disconnect appears to be exactly what's playing out here. The court found the underlying punishment illegal, yet the flag Michael described still stands, at least according to his own account days after the ruling.

That gap between legal findings and bureaucratic status is not just a technicality. For a company trying to sell services to the military, to intelligence agencies, or to contractors who depend on Pentagon approval, a lingering risk flag can chill business relationships even if the legal basis for it has collapsed. Investors, partners, and government buyers tend to move cautiously around ambiguity, and right now there's plenty of it.

It's worth pausing on why the Pentagon reportedly targeted Anthropic in the first place: criticism of the government. If that's accurate, and a federal judge has now said as much, it raises uncomfortable questions about how national security tools like risk designations can be used, intentionally or not, to punish dissent rather than to address genuine security concerns. Supply chain risk labels exist for good reasons. Foreign-owned components, unreliable manufacturing, or opaque ownership structures can all pose real threats to systems the military depends on. But when a label appears tied to political friction rather than technical vulnerability, it undermines the credibility of the entire vetting process, for this company and for every other firm relying on that same system to be fair and consistent.

The Defense Department did not respond to a request for comment on Michael's remarks or on how the designation squares with the court's ruling. That silence leaves a lot unresolved. Is the flag under review? Is it being maintained out of caution while the legal process plays out further? Or is there a genuine, separate security concern that hasn't been made public? Without an answer, outside observers are left to read tea leaves from dueling social media posts and cable news interviews.

This story is unfolding against a broader backdrop of the administration trying to balance two competing instincts: promoting American AI companies as a matter of economic and geopolitical strategy, while also maintaining strict controls over what touches military and defense systems. Lutnick's comments reflect the first instinct. Michael's reflect the second. Both are legitimate government priorities, but when they collide this visibly, it creates confusion for the companies caught in the middle and for the public trying to understand who's actually setting AI policy.

Anthropic has positioned itself as one of the more safety-focused AI labs, often emphasizing caution and oversight in its public messaging. Whether that reputation helps or hurts it here is an open question. A company known for pushing back on government requests might be more likely to end up on a risk list precisely because it doesn't simply comply. That's speculative, but it's the kind of speculation that fills the vacuum when officials won't explain their reasoning.

Why It Matters

This isn't just a dispute between one company and one agency. It's a preview of the friction the government will keep facing as AI firms become embedded in defense supply chains while also serving as public critics of policy. If a risk designation can outlast the legal justification for it, and if two top officials can't agree on where a major AI company stands, that's a signal the system for vetting these relationships needs clearer rules, not just louder statements.