Written by Sharon Chan
This article was first published truth
Anthropic’s CEO raised concerns about autonomous drones and the use of AI in surveillance.
Defense Secretary Pete Hegseth has threatened to blacklist Antropic if the AI company doesn’t allow its tools to be used for autonomous drone strikes and mass surveillance, a chilling demonstration of the Pentagon’s priorities.
In a meeting with the company on Tuesday, Hegseth said the company must lift the safety restriction requirement by Friday at 5:01 p.m. Otherwise, the Pentagon would declare the company a “supply chain risk,” effectively blacklisting it or, paradoxically, invoking the Defense Production Act to force Antropic to comply, officials warned.
Representatives from the company who attended the meeting expressed safety concerns about AI’s ability to reliably control weapons, according to a source familiar with the meeting. A lack of regulation over the use of AI in mass surveillance could also pose a risk, they said they told officials.
The company’s CEO, Dario Amodei, has repeatedly expressed concerns about these issues.
“You’re worried about swarms of autonomous drones, right? Constitutional protections in our military rely on the idea that there are humans who won’t obey unlawful commands. With fully autonomous weapons, there’s really no such protection,” Amodei said in an interview with podcaster Wes Ross.
Amodei also worries that AI could access and process private conversations captured by technology in people’s homes, which could be used to label people politically and “undermine” the Fourth Amendment.
However, after meeting with Hegseth, Anthropic announced that it would abolish its central safety policy, which puts guardrails on its AI development to reduce the risks AI poses to society. It’s unclear whether the change is related to the Pentagon’s request, but the timing has raised suspicions.
Legal experts say it’s unclear whether the Trump administration can use the Defense Production Act to force Anthropic’s hand.
Anthropic is negotiating a contract with the Department of Defense, and has reportedly previously offered to allow its AI systems to be used for missile and cyber defense. But the Pentagon insists the company must be allowed to use the tool for any military purpose.
The company’s AI model, Claude, was reportedly used by the Pentagon during the Caracas shelling that killed 83 people, including civilians, and during the operation to kidnap Venezuelan President Nicolas Maduro. a wall street journal The report, citing sources, said Claude was used by the Pentagon through a partnership between Anthropic and Palantir, which have contracts with the U.S. government.
Pentagon officials said in a statement that Hegseth’s request “has nothing to do with mass surveillance or the use of autonomous weapons,” but the Trump administration has persisted in overstepping its legal authority to subject Americans to more violence and surveillance.
“I want to be clear about what responsible AI means at the Department of the Army. Gone are the days of unbiased AI and other DEI and social justice infusions that constrain and confuse adoption of this technology,” Hegseth said in a talk at SpaceX headquarters in January. “We will not adopt AI models that make war impossible.”
Experts have warned that using AI models in warfare is dangerous. A recent study in which researchers pitted ChatGPT, Claude, and Gemini models against each other in 21 war scenarios found that in 95% of the simulation games, one of the models used nuclear weapons.
This article was originally published by Truthout and is licensed under Creative Commons (CC BY-NC-ND 4.0). Please maintain all links and credits according to republishing guidelines.
