‘incoherent’: Hegseth’s Anthropic Ultimatum Confounds Ai Policymakers 

Defense Secretary Pete Hegseth’s ultimatum to the artificial intelligence startup Anthropic is sparking shock and confusion among lawyers and AI policymakers, who accuse the Pentagon of making contradictory threats as it pressures the company to lift restrictions on the use of its powerful AI model.

Hegseth met with Anthropic CEO Dario Amodei on Tuesday to deliver a warning — give the military unfettered access to its Claude AI model by Friday evening or else have the government label it a “risk” to the supply chain. The designation, typically reserved for foreign firms with ties to U.S. adversaries, could ban companies that work with the government from partnering with Anthropic.

But Hegseth simultaneously threatened to invoke the Cold War-era Defense Production Act to compel Anthropic to work with the Defense Department and nix the company’s ethical red lines, which include restrictions on using Claude to surveil U.S. citizens or empower autonomous weapons. The government used the law during the Covid-19 pandemic to accelerate production of medical supplies and vaccines.

Dean Ball, a former AI adviser in the Trump administration, said the Pentagon is contradicting itself by forcing Anthropic to cooperate with the government even as it moves to label the company a security risk.

“You’re telling everyone else who supplies to the DOD you cannot use Anthropic’s models, while also saying that the DOD must use Anthropic’s models,” said Ball, who was the lead author of the White House’s AI Action Plan. He called it “incoherent” to even float the two policy ideas together, and “a whole different level of insane to move up and say we’re going to do both of those things.”

“It doesn’t make any sense,” said Ball.

The high-stakes standoff has its roots in the January operation to capture former Venezuelan President Nicolás Maduro. The U.S. military used Anthropic’s Claude AI model in the raid, The Wall Street Journal reported, prompting Anthropic to ask its corporate partner Palantir about the use of its product. Hegseth and other Pentagon officials bristled at those questions, eventually summoning Amodei to deliver this week’s ultimatum.

Last summer, Anthropic — along with its corporate rivals OpenAI, Google and xAI — signed a $200 million contract with the Pentagon.

In a January speech at SpaceX headquarters, a company owned by tech billionaire and xAI CEO Elon Musk, Hegseth declared that the DOD’s artificial intelligence “will not be woke,” and will operate “without ideological constraints that limit lawful military applications.” The comment was widely viewed as a shot at Anthropic, particularly since White House AI czar David Sacks has repeatedly criticized the company.

In a statement to POLITICO, DOD spokesperson Sean Parnell confirmed that the department’s relationship with Anthropic “is being reviewed.”

“Our nation requires that our partners be willing to help our warfighters win in any fight,” said Parnell. “Ultimately, this is about our troops and the safety of the American people.”

On Wednesday, Axios reported that the DOD has asked two top defense contractors to assess their use of Anthropic’s model as a step toward designating the company a supply chain risk.

A Pentagon official said Tuesday that the department plans to both label Anthropic a supply chain risk and invoke the DPA to compel it to work with the DOD unless it accedes to demands by Friday. A Pentagon spokesperson declined to comment on further questions about the strategy, including its alleged incoherence.

But Katie Sweeten, a tech lawyer and former Department of Justice official who served as the agency’s point of contact with the Pentagon, also called the DOD’s arguments “contradictory.”

“I don't know how you can both use the DPA to take over this product and also at the same time say this product is a massive national security risk,” said Sweeten. She warned that Hegseth’s “very aggressive” negotiating posture could have a chilling effect on partnerships between the Pentagon and Silicon Valley.

“This is about the heaviest-handed way you can regulate a business,” said Sweeten, calling it a “landmark moment with how the Department of [Defense] interacts with our cutting-edge technology developed on U.S. soil.”

During the Biden era, some conservatives objected to the Democratic president’s use of the DPA to force AI companies to share data on cutting-edge models. This time, voices from across the political spectrum are already decrying the Pentagon’s proposed use of the law to require a private company to lift restrictions on the use of its product.

In a Wednesday statement, Sens. Elizabeth Warren (D-Mass.) and Andy Kim (D-N.J.) criticized the administration's threat to use the Defense Production Act and warned it would “shatter the bipartisan consensus” in support of the statute.

On the other side of the aisle, Neil Chilson, head of AI policy at the Koch-affiliated Abundance Institute, warned that the Korean War-era emergency statute should not be used to expand executive power. Chilson called on Congress to change the law.

On Tuesday, Amodei reiterated his red lines on surveillance and autonomous weapons to Hegseth. An Anthropic spokesperson did not respond when asked whether the company will submit to the Pentagon’s demands, whether Anthropic plans to sue the Pentagon if it carries out its threats, or if the DOD wants to use Claude to surveil U.S. citizens or deploy weapons without humans in the loop.

A Pentagon official, granted anonymity to discuss sensitive talks, said in a statement that its standoff with Anthropic has “nothing to do with mass surveillance and autonomous weapons being used.”

The DOD official said other AI companies, including OpenAI, Google and xAI, “are working collaboratively with the Pentagon in good faith to ensure their models can be used for all lawful purposes.” The official confirmed to POLITICO that xAI has agreed to allow its AI model Grok to be used in a classified setting, and that OpenAI and Google are “close.”

It is unclear whether the other AI firms have agreed to allow the Pentagon to use their models to surveil U.S. citizens or empower autonomous weapons. Spokespeople for OpenAI, Google and xAI did not respond to questions on the subject.

Ball suggested the Pentagon is standing on “principle” and is not actually interested in crossing Anthropic’s red lines on surveillance or autonomous weaponry.

But Sweeten said it’s tough to understand why the Pentagon has launched such an unprecedented campaign against a U.S. tech company, unless it has a strong interest in using its tools in ways that Anthropic hopes to forbid.

“If these are the lines in the sand that the [DOD] is drawing, I would assume that one or both of those functions are scenarios that they would want to utilize this for,” said Sweeten.