U.S. military leaders, including Secretary of Defense Pete Hegseth, met with executives from artificial intelligence company Anthropic on Tuesday, intensifying a debate over what the government can do with the company’s powerful AI models. Axios reported that Hegseth ordered Anthropic CEO Dario Amodei to agree to the unit’s terms by Friday evening or face penalties.
Anthropic, which bills itself as the most secure of the big AI companies, has been mired in a weeks-long disagreement with the Pentagon over how the military should be allowed to use its large-scale language model, Claude. U.S. defense officials have called for unfettered access to Claude’s capabilities, but Anthropic is reportedly resisting allowing its products to be used in mass surveillance systems or autonomous weapons systems that can use AI to kill people without human intervention. The Department of Defense (DoD) has integrated Claude into its operations, but is threatening to sever ties with it over what its top brass perceives as an obstacle erected by Anthropic.
At stake in the negotiations is whether the AI industry will push back against government demands for military use of its products, a long-standing topic of debate among researchers and ethical AI advocates. Defense officials have already threatened to take punitive measures if Anthropic doesn’t comply, including terminating major contracts and designating the company as a “supply chain risk.”
Last July, the Department of Defense awarded contracts worth up to $200 million to several major AI companies, including Anthropic, Google, and OpenAI. But until this week, Anthropic’s Claude product was the only model approved for use in classified military systems. The Pentagon on Monday signed an agreement allowing military personnel to use Elon Musk’s xAI chatbot on classified systems, which has recently faced backlash for creating non-consensual sexual images.
According to the Washington Post, both xAI and OpenAI agreed to the government’s terms on the use of AI, and defense officials said OpenAI was allowed to use its models for “all lawful purposes.” OpenAI did not immediately respond to a request for comment on the agreement with the government.
Antropic’s meeting with the Pentagon comes a month after the U.S. military reportedly used Claude to help capture Venezuelan leader Nicolas Maduro. The Trump administration has widely pushed to incorporate AI into the military, and Donald Trump has repeatedly vowed that the United States will win the global AI arms race to dominate AI technology.
Emil Michael, the Pentagon’s chief technology officer and former Uber executive, has been publicly campaigning for Anthropic to “cross the Rubicon” and agree to the government’s terms.
“If someone wants to make money from the government or the U.S. Department of the Army, I think those guardrails should be tailored to our use case, as long as it’s legal,” Michael told Defense Scoop last week.
Meanwhile, Anthropic’s Mr. Amodei has long been vocal in support of stronger regulation of AI, and his company has supported political action committees that advocate stronger safeguards for artificial intelligence. Mr. Amodei opposes Mr. Trump in the 2024 U.S. presidential election, and Anthropic employs several former staffers of Mr. Biden, which was a factor in the pro-Trump venture capital firm pulling its investment in Anthropic earlier this year, the Wall Street Journal reported.
The Pentagon has spent billions of dollars in recent years pursuing AI-powered technologies, from unmanned drones to automated targeting systems. These technological advances are accelerating ethical questions about how much decision-making power to cede to AI when it comes to lethal force. With the fighting in Ukraine featuring deadly semi-autonomous drones that operate without human control, these arguments are no longer theoretical.
