The Trump administration has consistently criticized AI safeguards as “woke AI.” During a fiery address in January, Defense Secretary Pete Hegseth spoke out against AI systems with political constraints. “Department of War AI will not be woke,” he declared. “It will work for us. We’re constructing combat-ready weapons and systems, not chatbots for an Ivy League faculty lounge.”
With only a few days remaining before Hegseth’s stated deadline for Anthropic to abandon its apparently politically-motivated requirements for AI safety and assurances of nonmilitary use, Anthropic informed that it “continued good-faith conversations” with the Pentagon.
Anthropic confronts a deadline of 5:01 p.m. Friday to provide the Pentagon with full access to its AI technology or face blacklisting from the military supply chain, reported, as confirmed by the Associated Press.
This confrontation comes after prolonged discussions between the Defense Department and Anthropic regarding how the military can deploy the company’s AI. As reported, Hegseth cautioned Anthropic that the Pentagon could classify the company as a “supply-chain risk,” a classification typically used for hostile foreign companies, such as the Chinese-based , if the company fails to comply, compelling military contractors to sever ties with Anthropic. He also threatened to invoke the Defense Production Act, legislation the Trump administration utilized during the COVID pandemic to compel companies to increase production of medical supplies, a threat Hegseth reportedly repeated on Tuesday.
An Anthropic spokesperson stated that the company would assist government operations in accordance with the company’s guidelines for ethical AI, saying it will “continue to support the government’s national security mission in line with what our models can reliably and responsibly do.”
The dispute has evolved into a symbolic battle within a wider discussion about who should determine the conditions for AI use: technology companies or the U.S. government. The Pentagon granted Anthropic, along with , OpenAI, and xAI, agreements valued at up to $200 million last year. While until recently, Anthropic has been the sole AI company authorized for use by the Pentagon, the startup has maintained a strict position against military applications of its AI, banning its use in fully autonomous weapons and domestic surveillance. But Elon Musk’s xAI this week to let the Pentagon use its AI for classified systems, creating competition for Anthropic’s formerly exclusive relationship.
The Pentagon is said to have used Anthropic’s AI model Claude via its collaboration with Palantir during the American operation in Venezuela, which resulted in the apprehension of former Venezuelan President Nicolás Maduro. Anthropic subsequently contacted Palantir, inquiring about the deployment of its AI during the operation, which Palantir then alerted the Pentagon to, according to .
Loosening safety commitments
But Anthropic is gradually weakening its firm dedication to safety protocols. The AI company Tuesday unveiled a revised version of its Responsible Scaling Policy (RSP), originally published in September 2023, in order to stay competitive, indicating the updated policy responds to shifts in market conditions. “The policy environment has shifted toward prioritizing AI competitiveness and economic growth, while safety-oriented discussions have yet to gain meaningful traction at the federal level,” the read.
CEO Dario Amodei has indicated a possible relaxation of safety obligations, with podcast host Dwarkesh Patel that the company faces “commercial pressure” and that its rigorous safety protocols have restricted its capacity to compete with rivals operating under more relaxed guidelines.
In an exclusive interview with , Jared Kaplan, Anthropic’s chief science officer, stated the RSP modifications were driven by safety considerations, not competitive concerns. “We felt that it wouldn’t actually help anyone for us to stop training AI models,” Kaplan said. “We didn’t really feel, with the rapid advance of AI, that it made sense for us to make unilateral commitments … if competitors are blazing ahead.”