Defense technology moves away from Claude

Following the Trump administration’s Friday decision to blacklist Anthropic and designate its technology as a supply chain risk, defense technology companies are telling employees to stop using Claude and switch to other artificial intelligence models and assistants.

“Most of our companies are actively involved in large defense contracts, so their interpretation of the requirements is very strict,” said Alexander Harstrick, managing partner at J2 Ventures, which supports startups in the field.

In an email, Herstrick told CNBC that 10 of his portfolio companies that work with the Department of Defense are “in the active process of withdrawing the use of Claude from defense applications and replacing it with other services.”

Meanwhile, defense contractors are lockheed martin The company is expected to remove Anthropic’s technology from its supply chain, Reuters reported late Tuesday.

This is a sudden reversal for Anthropic, which derives about 80% of its revenue from corporate customers, CEO Dario Amodei told CNBC in January.

The company entered the Department of Defense ecosystem in late 2024 through partnerships with software and service providers. Palantir. Months after this agreement, the Claude became the first major model to be deployed in a classified government network through a $200 million contract with the Department of Defense. The popularity of this model continued to grow throughout the business world, especially in the field of coding assistants.

Secretary of Defense Pete Hegseth declared in X that contractors and suppliers doing business with the U.S. military will be prohibited from doing business with Anthropic.

The announcement came after Anthropic executives refused to comply with government demands over the use of its models. They wanted assurances that their AI would not be misused for fully autonomous weapons or for mass domestic surveillance of American citizens.

As CNBC previously reported, following the Trump administration’s announcement, Anthropic’s models are still being used to support U.S. military operations in Iran.

This problem is far from resolved.

Anthropic can still appeal through the legal system, but since nothing has officially happened, they haven’t taken any action yet. Primarily limited to posting on social media.

Anthropic said in a blog post Friday, citing federal law enacted by Congress, that Hegseth does not have the authority to restrict companies that work with Anthropic from doing business with the government.

If the supply chain risk designation were formalized, it would only apply when companies use Claude as part of a defense contract and “cannot influence how contractors use Claude to serve other customers,” the company wrote.

Anthropic declined to comment beyond mentioning the blog post.

But several defense technology executives, who requested anonymity due to the sensitivity of the matter, said they were preemptively moving employees away from Mr. Claude.

One defense executive told employees last week to switch Claude to other models, including some open source options, but said the process could take a week or two. This was in the run-up to Friday’s deadline, with both sides refusing to budge.

The CEO of another defense technology company announced this week that employees were instructed on Monday to stop using Claude until further guidance. The company should expect the ban to go into effect, the person said.

Herstrick, who served as a military intelligence officer in the Army Reserve and was deployed to Afghanistan and Iraq in 2017, said his company was making the switch “out of an abundance of caution.”

“This is in no way a reflection of Claude’s perceived shortcomings; most comments were that the product itself is excellent and the situation is deplorable,” he wrote.

Hours after the Pentagon’s announcement on Friday, OpenAI CEO Sam Altman said in a post on X that the company had agreed terms with the Pentagon regarding the use of its AI models. After coming under fire over the weekend, Altman followed up with a post on Monday admitting that his timing was “sloppy” and that the company “should not have rushed” the acquisition.

Altman posted an internal memo saying the company plans to amend the contract to include new language clarifying that “AI systems will not be intentionally used for domestic surveillance of U.S. persons and citizens.”

President Donald Trump said in a social media post Friday that federal agencies have six months to phase out the use of the technology. So far, in addition to the Pentagon, officials from the Treasury Department, the State Department, and the Department of Health and Human Services have told their employees to stay away from Mr. Claude.

One defense technology venture investor told CNBC that shutting down Anthropic won’t pose a big problem because serious companies doing business with the federal government will avoid relying on a single supplier.

Palantir, which relies on the government for nearly 60% of its U.S. revenue, declined to comment on its plans.

Analysts at Piper Sandler said in a note to clients on Tuesday that Anthropic is “heavily integrated into the military and intelligence communities” and that moving away from the company’s technology could be “short-term disruptive” to Palantir’s operations.

Analysts who rate Palantir stock as a buy write: “While AI capabilities can and will continue to be rebuilt as needed by new vendors, Anthropic has been a pioneer in operationalizing AI models for data-sensitive environments.” “Introducing and negotiating alternative technologies takes time and resources,” they write, which could be “spent on growth opportunities.”

Not everyone is in a hurry to act.

C3 AI Chairman and former CEO Tom Siebel counts the Department of Defense as a client and partners with consulting firms. booze allen hamilton. Mr. Siebel said in an interview that at this point, he does not see a “need to ease” Mr. Claude “until a lawsuit occurs.”

A partner at a defense-focused venture firm said his portfolio companies have limited exposure to Claude and are mostly users of OpenAI’s technology.

Tara Chiklovsky, CEO of Technovation, a global technology education nonprofit, said if the Pentagon pursues this strategy to the end and cuts off Anthropic, it could be a risky decision.

She said Anthropic is the most cautious modeler when it comes to building military systems, and that alternative suppliers used by the government are less secure.

The government also has a contract. google For use with Gemini and Elon Musk’s xAI for Grok.

“Once the dust settles, they will realize that Anthropic is the only company with this very unique skill set in technology,” Chiklovsky said. “Competition is so fierce that people think the only way to succeed is to ignore the weight of safety nets and go fast. Antropic shows that maybe that’s not the way to go.”

—CNBC’s Ashley Capoot and Jordan Novet contributed to this report

Attention: What is your red line?