Dispute centered on AI use in surveillance, autonomous weapons
Newly disclosed emails show in detail the uneasy back-and-forth between Anthropic CEO Dario Amodei and Undersecretary of Defense Emil Michael that led to the Pentagon labeling the AI company a supply-chain risk, according to court documents filed Tuesday.
The emails, part of ongoing litigation between the Defense Department and Anthropic, document how the relationship initially soured and offer a deeper look into the flurry of communication before the Pentagon effectively barred partners from using Claude.
The dispute centered on Anthropic’s opposition to the use of its models in domestic surveillance and AI-powered weapons. The Pentagon insisted it must have flexibility to make its own national-security decisions and should be able to use Anthropic’s model Claude as it saw fit.
Michael reached out to Amodei in January to check in after the holidays following weeks of delay in negotiations that had been ongoing since late last year. Amodei’s illness in December contributed to the setbacks.
Amodei attempted to clarify the company’s preference for bans on fully autonomous weapons and domestic surveillance. He focused on questions about how AI could aid the collection of data on Americans, referring to the Department of War, the unofficial name the Pentagon has revived.
Michael rejected Amodei’s proposals, saying the department insisted on using AI in all lawful-use cases. He said he did not want to “force anything unnatural” if the two sides were too far apart; Amodei later offered the same sentiment.
Amodei made another proposal to compromise on the weapons policy, but Michael pushed back, citing an in-person discussion with Tarun Chhabra, Anthropic’s head of national-security policy and a former Biden administration official.
Anthropic has disputed the Pentagon’s characterization of some of the meetings and discussions.
Amodei said following existing laws was inadequate because there are legal ways to use AI for surveillance. He cited a section of legislation that allows surveillance for foreign-intelligence purposes that sometimes includes communications of Americans.
Michael defended the Pentagon’s practices, saying they followed the law and that Anthropic in some cases was asking for language and compliance processes that went beyond what the Defense Department could implement.
Amodei rejected the Defense Department’s latest proposal, which he said added loopholes, and conceded there was not a path forward to working together.
The following day, Defense Secretary Pete Hegseth said on social media that he was directing his department to label the company a supply-chain risk, which meant businesses working with the Pentagon could no longer use Anthropic’s AI models. Anthropic has sued over the designation.
The same day Hegseth targeted Anthropic, President Trump directed agencies to stop using the company’s tools. A federal judge ordered the Trump administration to desist from applying the president’s directive and from implementing aspects of the supply-chain risk designation. The government is appealing while a lawsuit over other aspects of the designation plays out.
The documents also show the race by federal agencies to remove Claude from their systems. On March 3, shortly after 10 p.m., Commerce Department official Olivia Bradley emailed colleagues at the agency.
Michael recently said that two-thirds of the operations in the Pentagon that were using Anthropic’s models have switched to other AI tools.
“While this case was necessary to protect Anthropic, our customers, and our partners, we remain focused on working productively with the government to ensure all Americans benefit from safe, reliable AI,” an Anthropic spokeswoman said.