Secretary of War Pete Hegseth is threatening to end artificial intelligence company Anthropic’s $200 million contract and “blacklist” it, revealed sources who spoke to multiple news outlets.
Two “sources familiar with the matter,” told CNN that the Pentagon wants Anthropic to lift its restrictions when used for the military. According to the outlet, Hegseth gave Anthropic CEO Dario Amodei a Friday deadline for compliance during a Tuesday meeting.
“He also threatened to put the AI company on what could amount to a government blacklist,” said CNN.
Anthropic was founded in 2021 by people who previously worked at another AI company, ChatGPT maker OpenAI and left it over disagreements related to safety, CNN noted. On its website, Anthropic describes itself as a “safety first” company.
Even with this safety-first approach, AI use comes with risks, Bloomberg’s Rachel Metz recently explained in an interview with Holly Quan from Audacy station KCBS in the Bay Area.
“There are definitely a lot of security risks in general when you are allowing an AI program to take over your computer or your internet browser,” Metz said. “One really commonly known security concern is this idea of prompt injection attacks. And that’s where an AI model could be manipulated by a malicious command.”
Maintaining security can be tricky in a world with rapidly developing technology. Editor-in-Chief Jeffrey Goldberg of The Atlantic revealed last year that Hegseth accidentally texted him war plans via the Signal messaging app, for instance.
Anthropic also has concerns over two issues “that it isn’t willing to drop,” per CNN’s source. These are the potential for AI-controlled weapons and use of AI for mass domestic surveillance of American citizens.
“According to one source familiar, Anthropic believes AI is not reliable enough to operate weapons, and there are no laws or regulations yet that cover how AI could be used in mass surveillance,” the outlet said.
It said negotiations between the Pentagon and Anthropic have been going on for months. POLITICO reported that the company’s ties came under scrutiny recently, following a report that indicated its Claude AI model was used by the U.S. military to capture former Venezuelan President Nicolás Maduro.
Axios first reported details of the Friday deadline for Anthropic to dial back its safeguards for the military. Both CNN and POLITICO said a Pentagon official confirmed that Anthropic now has until 5:01 p.m. Friday to comply and said that the issue doesn’t have anything to do with AI being used for weapons or mass surveillance.
From there, things get a bit confusing. CNN reported that the official said Hegseth plans to use the Defense Production Act on Anthropic, essentially “compelling them [Anthropic] to be used by the Pentagon regardless of if they want to or not.”
This law allows the government to influence businesses in the interest of national defense and it was used by both Trump and former President Joe Biden during the COVID-19 pandemic. POLITICO also noted that “some conservatives objected to the Biden administration’s use of it to force AI companies to share data on cutting-edge models with the government,” in 2024.
At the same time, the official also said that Anthropic would be labeled a supply chain risk. That would “prohibit companies with military contracts from using Anthropic’s products in any of their military work,” explained.
“Katie Sweeten, a former liaison for the Justice Department to the Department of Defense, said she’s not sure how the Pentagon can both declare a company to be a supply chain risk and compel that same company to work with the military,” said CNN.
She said: “What it sounds like is that the supply chain risk may not be a legitimate claim, but more punitive because they’re not acquiescing.”
As for Antrhopic, CNN reported that “a source familiar” said the company has no plans to adhere to the Pentagon’s demands.
According to the outlet, “it could deal a major blow to the AI firm at a time when it’s trying to expand its reach in the enterprise space,” if the Pentagon follows through with the alleged plans on Friday.