On September 23, Washington offered two answers to the same question, whether and how to rein in artificial intelligence, and they pointed in opposite directions.
In the Senate, Bernie Sanders formally introduced legislation that would ban the development of artificial superintelligence and create a new federal department to oversee the technology. Hours later, from the podium of the United Nations General Assembly, Donald Trump rejected AI regulation outright and proposed something narrower: that the country stop calling the technology AI and use the word “superintelligence” instead.
Sanders, working with Representative Greg Casar, has been building toward the bill for weeks. The measure, the Ban Artificial Superintelligence Act, would prohibit any person or entity from developing or deploying systems that exceed human control, pause development of the most advanced models, and establish a Department of Artificial Intelligence to write and enforce the rules. It is the sharpest regulatory proposal yet from a senator who has spent the year warning about the technology’s risks.
The Vermont senator’s record gives the bill weight beyond its slim chance of passage. Sanders was an early advocate of pausing advanced AI development, and he pushed a nationwide moratorium on new data centers, a position that has since drawn bipartisan interest as the energy and water demands of AI have collided with local politics. People familiar with the drafting said the bill is aimed as much at shifting the debate as at becoming law.
Sanders has been moving toward this position for more than a year. The data-center moratorium he championed argued that the physical footprint of AI, the power plants and water pipelines it requires, was growing faster than communities could absorb. The superintelligence bill extends that logic from the footprint to the technology itself, arguing that a system capable of exceeding human control should not be allowed to exist regardless of who builds it.
The bill’s definitions are deliberately broad. Superintelligence, in the draft, means a system that exhibits or can readily be induced to exhibit capabilities that outpace human control across a wide range of tasks, language that leaves enforcement to a department the legislation would create. The new Department of Artificial Intelligence would license, monitor and restrict development, a structure modeled on agencies that regulate other consequential technologies.
Trump’s answer came from a different premise. In his address to the General Assembly, he argued against AI regulation on the grounds that it would hand an advantage to rival nations, and he floated “superintelligence” as the preferred name for the technology. The rebrand, in his telling, was about framing: the word “artificial” undersells what the systems are becoming, while “superintelligence” captures their power without inviting the caution that has attached itself to AI.
The speech, delivered to an audience of world leaders, framed the technology in the language of national advantage. Trump argued that slowing development would cede ground to China and that the United States should lead rather than regulate. The suggestion to retire the term “artificial intelligence” was characteristic, a branding move wrapped around a policy position, and it landed with the force of a presidential preference even though it carried no legislative weight.
The two statements, delivered within hours of each other, frame the divide now running through American AI politics. One side sees machines that could outpace human control and wants to prohibit them before they exist. The other sees a technological race that the United States can win only by refusing to slow itself down. Between them sits a Congress that has spent the year debating narrower measures, export controls, energy policy, safety reporting, without settling the underlying question.
Sanders’ proposal is unlikely to pass in anything like its current form. A permanent ban on a technology that the country’s largest companies are racing to build would face opposition across the industry and from most of the Senate’s own membership. But the bill’s function may be less legislative than rhetorical, a marker of how far the safety argument has traveled from the fringes toward the center of the debate.
The same could be said of Trump’s rebrand. Renaming a technology does not change what it does, but names shape regulation, funding and public fear. The fight over whether to call these systems artificial intelligence or superintelligence is, in effect, a fight over how alarmed the country should be, and both men understand that the word chosen matters almost as much as the policy attached to it.
The split is not merely rhetorical. It maps onto a genuine disagreement about what the technology is. Sanders’ camp treats systems that could outpace human judgment as an existential risk that should be preempted. Trump’s camp treats the same systems as an economic and military race that the country can win only by running faster. Both views have constituencies, and neither has a governing majority, which is why the day produced a bill that will not pass and a speech that will not bind.
What the day made clear is that the United States has not settled on a single story about the technology it is building. Sanders wants the state to step in before the machines outrun their makers. Trump wants the state to step back and let the machines be built under a more confident name. The distance between those positions, measured over a single September day, is the distance the country still has to travel before it has anything resembling a policy.
