The Pentagon’s growing grip on AI

According to Brookings Institution’s Thomas Wright, pressure from the Pentagon could make tech companies more cautious about working with the government. If companies take the wrong step, he warns, officials could “either partially nationalize your company or… ruin your company and burn it to the ground by designating you as supply chain risk.”

Wright’s comments come after Anthropic got into a dispute with the Defense Department late last month. Washington wanted to have unfettered access to its AI systems, but CEO Dario Amodei refused this request. Subsequently, the Trump administration labeled Anthropic a “supply chain risk,” prompting Anthropic to sue the government, in turn.

Wright’s view: AI systems should still have a human safeguard. He explains that means keeping a person “on the loop” and able to intervene or shut down a weapon system if necessary.