Casar and Sanders want to ban superintelligence in the US; the White House is already saying no

By Zendoric · September 25, 2026.

Democratic Representative Greg Casar (Austin, Texas) and Senator Bernie Sanders have introduced a bill to ban the development of "artificial superintelligence," defined in the text as a system with cognitive capabilities that surpass those of humans and with the potential to "destroy humanity." The measure would also create a cabinet-level federal agency tasked with overseeing AI systems and pushing for international agreements to curb the technology in other countries, according to Spectrum News.

The timing is no accident. The initiative comes as President Donald Trump meets with Chinese President Xi Jinping, with AI safety on the bilateral agenda. Trump, however, has already staked out his position: in his speech to the UN General Assembly this week he said he does not intend to "slow down the growth of something that will be bigger than the industrial revolution," and on social media he has called for leaving superintelligence "exactly where it is," with the Justice Department as the only control mechanism. The White House, through deputy press secretary Liz Huston, went further and accused Democrats of having applauded a Biden-era approach that, she said, "stifled innovation" and "hurt consumers."

Casar makes the case for urgency with a deliberately alarming image: "The AI arms race is not a race you win. It's the kind of race where, when you cross the finish line, you fall off a cliff." Brendan Steinhauser, a Texas political strategist and director of the Alliance for Secure AI—the organization calling for congressional action on the issue—frames the risk in very concrete terms: cybersecurity, critical energy and water infrastructure, and "loss-of-control scenarios." Steinhauser insists this "is not a partisan issue" and welcomes the fact that it is precisely two of the most progressive lawmakers in Congress who have put it on the table.

It is worth separating here two levels that the article itself mixes together. One is the risk diagnosis—loss of control, critical infrastructure, weapons of mass destruction—which matches the concerns that safety researchers, and even some of the leading labs themselves, have been voicing for months. The other is the political viability of this particular bill, which is practically nil: with a Republican-majority Congress and a White House that has already branded any attempt at pre-emptive regulation "stifling," this has all the hallmarks of a messaging bill, designed to stake out a position and spark debate, not to advance through the legislative process.

Our reading is that the episode matters less for the bill's legislative fate than for what it reveals about the state of the public conversation in Washington. Banning by law something that does not yet even exist—there is no technical consensus on when, or whether, a real superintelligence will be reached—is a pre-emptive regulatory bet, similar in spirit to nuclear non-proliferation treaties: the aim is to set the limit before the capability exists, not after. It is a defensible approach in theory, but it runs into a problem we have flagged in other analyses: labs' self-regulation cannot be verified from the inside, and a federal agency only works if it has independent auditing teeth, not if it merely "monitors." The text, as the source describes it, is still short on detail about how that oversight would be enforced or what the consequences of breaching it would be.

Over the longer term, the episode confirms a tension that will shape the coming years: the Trump administration has decided to bet on speed against China as its number one priority, and any domestic regulatory brake—wherever it comes from—is read as a threat to that race. It is an understandable reading from a geopolitical standpoint, but also a simplification: winning the technology race and managing its catastrophic risks are not incompatible goals, they are two halves of the same problem. Our underlying thesis does not change because of this episode: AI can take us to a radically better society of abundance and health, but only if someone sees to it, in parallel, that the road there does not run through hackable critical infrastructure or systems out of control. Bills like Casar and Sanders's, even if they go nowhere, keep that conversation alive at a moment when the White House would rather not have it.