Anthropic has been providing services to the U.S. government through Palantir Technologies for more than a year. Palantir is an enterprise software company known for its work with the Pentagon and other federal agencies. Palantir's customers use Anthropic's models to identify patterns in massive amounts of confidential data to aid decision-making.

Anthropic CEO Dario Amodei and Palantir CEO Alex Karp.
Anthropic CEO Dario Amodei and Palantir CEO Alex Karp.

But on Friday, the U.S. Department of Defense made a decision to restrict the use of Anthropic’s technology by military contractors, which may force Palantir to cut ties with the AI ​​company. Palantir is one such contractor: About 42% of its nearly $4.5 billion in revenue last year came from U.S. government contracts, company disclosure documents show.

If Palantir decides to stop using Anthropic in its federal business, it would cut off one of the AI ​​startup's revenue streams -- government customers paying to run their models in Palantir's core analytics and database products. While those sales don't contribute much to Anthropic's projected $18 billion in revenue this year, Palantir's status as a favored government contractor makes it a premium partner for software companies, especially those with large customers.

According to people familiar with Palantir's product operations, Palantir can switch to other AI model suppliers and its revenue from related contracts will be largely unaffected.

Palantir provides customers with customized software and AI applications, some of which are specifically designed to fit Anthropic’s Claude model. This means that if Claude is abandoned, Palantir will need to adjust its services, which may take several weeks, the person said.

On Tuesday, Palantir CEO Alex Karp appeared to criticize Anthropic without naming it directly. Speaking at the Defense Technology Summit hosted by Andreessen Horowitz in Washington, he railed against Silicon Valley for working against the U.S. military and warned that AI companies could anger both liberals and conservatives.

“It would be foolish for Silicon Valley to think that we’re going to take away everyone’s white-collar jobs... and undermine the military, but not think that’s going to lead to the nationalization of our technology,” Karp said. “That’s where this is going.”

On Friday, Defense Secretary Pete Hegseth said he had directed the government to classify Anthropic as a "supply chain risk" after a dispute with Anthropic over restrictions on the military's use of its AI. Hegseth said the directive would ban any company that does business with the U.S. military from doing business with Anthropic — a potentially harsh directive that could theoretically cut off the startup from cloud service providers Amazon and Google that also have defense contracts.

After negotiations broke down, OpenAI quickly reached an agreement with the Pentagon to have its models take over classified military work, ostensibly replacing the military's use of Anthropic's models.

When the Department of Defense formally designates Anthropic, the language may not be so broad. As of Monday evening, the Defense Department had not issued a formal designation, a process that could take months. Legal experts say the process includes conducting a risk assessment, notifying Congress and considering less intrusive ways to reduce supply chain risks.

U.S. President Donald Trump posted on the "Real Social" platform last Friday that agencies such as the Department of Defense that use Anthropic products will have a "six-month phase-out period."

Anthropic CEO Dario Amodei argued that the designation “applies only to the use of Claude as part of a Department of War (the Trump administration’s preferred term for the Department of Defense) contract.” Legal experts say Anthropic can fully fight the designation, in part because the law generally applies to non-U.S. companies.

Anthropic also said on Friday it would take legal action in court over the "supply chain risk" determination. Anthropic has not yet filed a lawsuit because the designation has not officially taken effect, according to people familiar with the decision.

Anthropic said it "strongly prefers to continue to provide services to the Department of Defense and the warfighter - subject to the implementation of the two safeguards we proposed," suggesting that negotiations may still be ongoing if the two sides can reach an agreement.

Anthropic sought to reassure customers and business partners, saying the Defense Department's decision, if implemented, would only limit the use of its technology by government contractors in work with the U.S. military, not all companies working with the federal government.

According to people familiar with Palantir's products, Palantir currently allows customers to choose to use AI models from providers such as Anthropic, OpenAI, Google and others when analyzing the data collected by its platform. The Pentagon has been using Anthropic models hosted by Amazon Web Services and run with Palantir software, which the companies previously said allowed U.S. defense and intelligence agencies to "make more informed decisions in time-sensitive scenarios while retaining decision-making autonomy."

In addition to the Anthropic model, the military already uses models from OpenAI, Google and other vendors through Palantir software, according to people familiar with the Pentagon's product use.

As of this week, employees at tech companies that do business with the government said federal contractors are awaiting guidance from the government on whether and how to cut ties with Anthropic to keep federal contracts.

Anthropic's direct work with the Pentagon has overshadowed its contract with Palantir over the past week, but previous collaborations with Palantir have also raised concerns within Anthropic.

As Anthropic finalized its partnership with Palantir in the fall of 2024, some Anthropic employees raised concerns internally about the way government agencies were being allowed to use its models, according to two people familiar with the matter. To allay those concerns, members of Anthropic's policy team hosted a question-and-answer session for employees, they said.