The conversation about artificial intelligence is shifting with a speed that has left even seasoned observers dizzy. For years, the default assumption in Washington and Silicon Valley was that frontier AI would remain in private hands, lightly regulated and driven by market incentives. That assumption is crumbling. From the democratic socialist left to the populist nationalist right, a new consensus is hardening: artificial intelligence is too powerful, too central to national security and economic fairness, to be left entirely to corporations. The public will own it. The question that no one is answering clearly is who exactly “the public” is, and who will hold the reins.
The most vivid accelerant of this debate arrived in the form of Alex Karp, the eccentric and fiercely patriotic CEO of Palantir Technologies. Speaking at the South by Southwest conference in March 2024, Karp issued a prophecy that sounded more like a threat. “I think the reality is that the government is going to have to step in and nationalize the AI space to some extent,” he told the audience. He did not say “regulate.” He did not say “partner.” He said nationalize. In follow up interviews, Karp made clear that he believes the creeping nationalization will be comprehensive. Not a stake. Not a board seat. All of it.
What makes Karp’s forecast more than a headline is the sustained campaign that accompanied it. According to multiple reports, the Palantir chief spent six months prior to his public remarks delivering private warnings to the most powerful AI executives in the country. In closed door dinners and one on one calls, Karp repeatedly told them that their industry was on a collision course with the federal government. Their refusal to fully embrace defense and intelligence work, their public agonizing over military contracts, and their insistence on treating AI as a pure consumer product were, in his view, building an unstoppable case for state takeover. He told them, essentially, that they could either serve the nation willingly or watch the nation seize their technology by force of law. The six month warning period is now over, and Karp’s private message has become a public gauntlet.
While Karp was making his rounds in Silicon Valley, a parallel argument was being built on Capitol Hill. Senator Bernie Sanders, the longtime champion of public ownership in essential sectors, turned his attention to artificial intelligence with a demand that shocked even his allies. In a series of Senate hearings and public statements throughout 2024, Sanders began arguing that the federal government should immediately acquire a 50 percent equity stake in the major AI corporations. His reasoning was characteristically blunt. AI, he argued, is not an ordinary technology. It is a general purpose invention on the scale of electricity or the steam engine, and its gains will concentrate so rapidly that society will fracture unless the public intercepts a massive share of the value. A 50 percent stake, according to Sanders, would ensure that the profits from automation fund universal healthcare, free higher education, and a robust safety net for displaced workers. It would also give the public a controlling interest in the safety and alignment of the most powerful systems ever created. In Sanders’s vision, 50 percent is the floor, not the ceiling.
The third force in this extraordinary alignment comes from the political figure who has spent a career attacking Sanders’s worldview. Donald Trump, the former president and current Republican nominee, has begun outlining his own version of public AI ownership. On the campaign trail and in a series of posts on his social media platform, Trump has repeatedly called for a national artificial intelligence initiative on the scale of the Manhattan Project. His language is rooted in the great power competition with China. He describes a government built AI infrastructure that would be “owned by the American people” and used to supercharge military superiority, energy independence, and domestic manufacturing. While Trump’s framing is nationalist rather than socialist, the destination is strikingly similar. Both Sanders and Trump are now arguing that the most advanced AI must belong to the state, not to a handful of California companies.
The convergence of these three voices signals a profound shift in the Overton window. It was not long ago that the mere suggestion of government ownership in a high tech sector was dismissed as fringe. Today, a defense industry billionaire, a socialist senator, and a populist former president are all, in their own ways, fighting for the same banner: public control of the means of intelligence. The reasons diverge. Karp sees nationalization as a necessary tool for national security. Sanders sees it as a tool for economic justice. Trump sees it as a weapon in a civilizational struggle. But the policy prescription is a radical departure from the status quo, and it is no longer a theoretical discussion. It is a live political project.
The entire debate now hinges on a single, terrifyingly open question. Who controls that future? The word “public” in public ownership is doing an enormous amount of work, and it conceals a chasm between competing models of government control. A nationalized AI sector could mean a transparent, democratically accountable agency governed by a bipartisan board, subject to the Freedom of Information Act, and designed to serve the broad public interest. Or it could mean an executive branch instrumentality under the direct control of the president, with minimal congressional oversight, embedded inside the national security apparatus and shielded from public view. The difference is the difference between a public library and a secret police archive.
Sanders has begun sketching his model. It involves a Public AI Administration modeled loosely on the Tennessee Valley Authority, an independent public corporation that would hold the government’s 50 percent stake and distribute dividends directly to citizens. Its board would include labor representatives, civil rights advocates, and scientists. Its meetings would be open. Its algorithms would be audited. Trump’s vision, to the extent it has been articulated, is almost the inverse. His Manhattan Project for AI would be run from the White House, closely integrated with the Department of Defense, and optimized for speed and secrecy. Public ownership under that framework could quickly become indistinguishable from a state owned enterprise in a command economy, with all the attendant risks of surveillance, political bias, and the weaponization of intelligence against domestic opponents.
Then there is the Karp factor. Alex Karp does not make predictions from a neutral perch. Palantir is already the de facto data operating system for large swaths of the U.S. military and intelligence community. If full nationalization arrives, Palantir is uniquely positioned to become the prime contractor, the default interface between the government and the nationalized AI infrastructure. Karp’s six month warning tour can be read, generously, as a patriotic wake up call. It can also be read, less generously, as a sophisticated lobbying campaign designed to accelerate the very outcome that would enrich his company and lock in its centrality for decades. The fact that both interpretations can be true simultaneously is part of what makes the current moment so disorienting.
The tech industry’s response to this gathering storm has been a mixture of denial and panic. The largest AI labs, many of which have spent years trying to establish ethical charters and voluntary safety commitments, are now confronting the possibility that their entire ownership structure could be overturned by a political alliance that spans the left and right. Some executives are quietly exploring defenses, including the relocation of critical assets overseas or the creation of complex trust structures designed to frustrate a government takeover. Others are belatedly trying to make the kinds of patriotic commitments that Karp demanded months ago, hoping that a credible embrace of defense work will slow the march toward nationalization. It may already be too late. The political logic that connects Sanders’s democratic socialism to Trump’s America First nationalism does not depend on the goodwill of tech CEOs. It feeds on their perceived arrogance and their distance from ordinary voters.
What makes the moment especially volatile is that neither the public nor the political class has fully internalized the velocity of AI progress. The systems being built today are on a trajectory toward artificial general intelligence that may be just a few years away. Once an AI system can outperform human experts in the design of advanced weapons, the management of financial markets, and the manipulation of information ecosystems, the argument for public control will become overwhelming and bipartisan overnight. Karp’s prediction of full nationalization is not a distant science fiction scenario. It is an extrapolation from capabilities that are already in the pipeline. The six months of private warnings he issued are likely a preview of a much larger shift that will unfold in the next presidential term.
In that light, the question of who controls the nationalized AI future is not a secondary detail. It is the constitutional fight of the century. Congress is currently asleep, with no serious legislative framework for public AI ownership on the table. The executive branch is amassing power through procurement contracts and emergency authorities that could easily be converted into ownership stakes without a single vote. The judiciary has barely begun to think about what the Fifth Amendment’s takings clause means when the asset being taken is a model weight file or a training dataset. The institutions that would be needed to administer a democratic form of public ownership do not exist yet, and no one is building them.
If Alex Karp is right, if total nationalization is coming, then the window for designing the control structure is closing fast. The difference between Bernie Sanders’s 50 percent public stake and Donald Trump’s 100 percent state AI machine is not a matter of degree. It is a fork in the road of American governance. One path leads to a publicly accountable, egalitarian infrastructure that could fund a new social contract. The other leads to a concentration of technological power in the hands of a single leader, accountable only to the next election and his own instincts. Both paths are now being paved. The campaigns, the think tanks, and the private dinners are already mapping them out.
Karp’s public prophecy has achieved one undeniable effect. It has forced the conversation out of the academic journals and into the open. Sanders has named a number. Trump has sketched a vision. Karp has predicted the whole thing. The only missing piece is a sustained, democratic debate about what the word “our” actually means when we say the AI should belong to the public. Because the public is not a monolith, and the state is not a neutral vessel. The battle for AI ownership will not be won by the side that shouts loudest. It will be won by those who show up to write the laws, the treaties, and the institutional blueprints before someone else does it for them. The countdown has started.
