Palantir’s Alex Karp warns frontier AI labs may face nationalization

Alex Karp, the CEO of a company that makes billions selling software to governments, now thinks the government should own the companies building the most powerful AI systems. It’s a position that sounds radical until you hear his reasoning, which is less about ideology and more about who’s going to foot the bill when things go wrong. In a CNBC interview on September 17, 2026, Karp laid out a blunt case: frontier AI labs like Anthropic face liability exposure so massive that no private company can realistically manage it. The lawsuits from clients alone, he argued, could be existential. And the only entity with the capacity to cap that kind of exposure is the US government. The liability problem no one wants to talk about Karp’s argument hinges on a deceptively simple observation. When an AI system causes harm, whether through faulty outputs, intellectual property violations, or something worse, somebody has to pay. Right now, the companies building these systems are on the hook. Rather than a government backstop, he’s suggesting the government should simply own the labs outright. In his telling, nationalization isn’t a punishment. It’s the only practical solution to a math problem ...