Anthropic, AI company that helped US capture Maduro rejects Pentagon 'threats' and invokes 'democratic values'

The problem for Anthropic is the potential use of its tools, such as its AI assistant Claude, for two purposes: “mass domestic surveillance” of citizens and “fully autonomous weapons,” Reuters and the BBC write. The Pentagon insists it doesn't want that, but demands that Anthropic remove any restrictions on its AI systems, which the company refuses.

The CEO of Anthropic said Thursday that it will not budge in its dispute with the Pentagon over how its artificial intelligence (AI) technology is used.

Dario Amodei said his firm would rather stop working with the Pentagon than accept the use of its technology in a way that could “undermine, rather than defend, democratic values.”

Amodei's comments come two days after a meeting with US Defense Secretary Pete Hegseth in which he demanded that Anthropic accept “any lawful use” of its tools. The meeting ended with a threat to remove Anthropic from the Pentagon's supply chain.

“These threats do not change our position: we cannot accept their request with a clear conscience,” Amodei said.

Reuters writes that Amodei's position calls into question the $200 million contract with the US government.

The reason for the conflict between Anthropic and the Pentagon

The problem for Anthropic is the potential use of its tools, like its AI assistant Claude, for two purposes: “mass domestic surveillance” of citizens and “fully autonomous weapons.”

Amodei stated that “such use cases have never been included in our contracts with the War Department and we believe they should not be included now.”

“Should the Department decide to move away from Anthropic, we will work to ensure a smooth transition to another provider,” Amodei said.

An Anthropic spokeswoman added Thursday that while the company received an updated version of the contract from the Defense Department late Wednesday, it offers “virtually no progress toward preventing Claude from being used for mass surveillance of Americans or in fully autonomous weapons.”

“The new wording presented as a compromise was accompanied by legal terms that would allow these safeguards to be ignored at will,” she said.

“Despite (the War Department's) recent public statements, these limited guarantees have been the focus of our negotiations for months,” she added.

The Pentagon's response

Emil Michael, the US undersecretary of defense, personally attacked Amodei on Thursday night, writing on X that he “wants nothing more than to try to personally control the US military and is willing to jeopardize the security of the country.”

“It is a shame that Dario Amodei is a liar and has a God complex,” the official added, assuring that the Pentagon “will always respect the law, but will not bow to the whims of any for-profit technology company.”

Also Thursday, Pentagon spokesman Sean Parnell told X that the department has no interest in using AI for mass surveillance of Americans or developing autonomous weapons that operate without human involvement.

A Pentagon official previously told the BBC that if Anthropic did not comply, Hegseth would ensure that the Defense Production Act was invoked against the company.

This law essentially gives the US president the authority to deem a particular company or its product so important that the government can require it to meet defense needs.

Hegseth also threatened to label Anthropic a “supply chain risk,” which would compromise any contracts it has with the US government.

A former Defense Department official, who asked not to be named, told the BBC on Thursday that Hegseth's reasons for both measures were “extremely flimsy”.

How Anthropic's CEO explains his refusal

A person familiar with the negotiations, who asked not to be named, said that tensions between Anthropic and the Pentagon “date back several months” before it became public knowledge that Claude was used in the operation to capture Venezuelan President Nicolás Maduro, writes the BBC.

While Amodei didn't specify exactly how Anthropic might be used by the Pentagon for mass surveillance or fully autonomous weapons, he wrote in a company blog post that AI can be used to “assemble scattered, individually innocuous data into a comprehensive picture of anyone's life — automatically and at scale.”

“We support the use of AI for legitimate foreign intelligence and counterintelligence missions,” Amodei said. “But using these systems for mass domestic surveillance is incompatible with democratic values,” he added.

Regarding the use of AI in weapons, Amodei stated that even the most advanced and powerful AI systems today “are not reliable enough to power fully autonomous weapons.”

“We will not knowingly supply a product that endangers American soldiers and civilians,” Amodei said.

“Without proper oversight, fully autonomous weapons cannot be trusted to exercise the critical judgment that our professional, highly trained troops demonstrate every day. They must be used with adequate safeguards, which currently do not exist,” Amodei said.

He added that Anthropic “offered to work directly with the War Department on research and development to improve the reliability of these systems, but the offer was not accepted.”