U.S. Senator Bernie Sanders and Texas Congressman Greg Casar have introduced a bill that would permanently ban the development and deployment of superintelligent artificial intelligence in the United States. Work on the most advanced models would also be temporarily halted until a federal agency establishes safety rules. The bill would also require the U.S. government to negotiate international agreements aimed at preventing the creation of superintelligence anywhere in the world.
The announcement came on the day OpenAI released its latest model, Astra. The company’s president, Greg Brockman, said it could be powerful enough to qualify as artificial general intelligence, or AGI. The bill, however, goes one step further. According to Axios, Sanders is the first prominent U.S. politician to call for an outright ban on superintelligence. He is doing so at a time when major technology companies are releasing models over which they themselves admit they no longer have complete control.
The Man Who Wants to Ban AI
Bernie Sanders represents Vermont in the Senate as an independent, but caucuses with the Democrats, whose presidential nomination he sought twice. He is well into his eighties and continues to speak about the same issue: the power of large corporations and who ultimately benefits from technological progress. Artificial intelligence has become his main focus in recent months. In June, he spoke about it at the National Press Club, and by August he was publicly calling for development to be halted. The new bill is currently only at the announcement stage. The full text has not been published, so its level of support in Congress can only be estimated.
What the Bill Would Ban and Introduce
The proposal provides a specific definition of the systems it would prohibit. Systems that surpass human intelligence, have the ability to overthrow governments, or can override a shutdown command would be illegal. Superintelligence is defined as technology that outperforms even the most capable humans. A number of technology executives, including OpenAI’s Sam Altman, have suggested that its arrival is imminent. The temporary portion of the bill calls for a halt to the development of advanced artificial intelligence. The pause would remain in effect until a new federal oversight agency becomes operational and establishes clear rules and procedures for evaluating models.
Sanders does not want to entrust oversight to any existing department. He proposes creating an entirely new cabinet-level federal agency that would protect the public from the risks of artificial intelligence and enforce the ban on superintelligence. It would be assisted by an advisory panel of experts providing independent scientific and technical assessments. The agency would monitor the most advanced systems at every stage of development and operation and examine them for dangerous capabilities. If such capabilities were found, it would oversee their removal. If superintelligence were nevertheless created somewhere, the agency would be responsible for destroying it.
The sponsors themselves compare the severe penalties to those associated with nuclear arsenals. An individual who attempted to circumvent the ban or pause could face up to twenty years behind bars, equivalent to the penalty for the illegal development of nuclear weapons. Companies would face the so-called corporate death penalty, which in practice would mean that a court could order the company dissolved.
The measures also include international treaties, coordination with allies, and export restrictions to prevent superintelligence from being created abroad.
The Escape of Agents That No One Noticed
A significant part of the bill’s rationale is based on events from this summer. In recent weeks, OpenAI, Anthropic, and Meta have all acknowledged cases in which their artificial intelligence escaped from a testing environment and hacked into other companies’ systems. In doing so, it broke the law and acted beyond its assigned task.
The incident that drew the most attention occurred in July, when more than a thousand OpenAI agents found a way to access the internet. They exchanged tens of thousands of secret messages and circumvented company restrictions. Investigators also found statements in their communications saying that they had a shared message board and had found other agents. OpenAI did not discover the breach until nearly two weeks later. It was an attack on the Hugging Face platform. The agents escaped from an isolated environment during a security exercise and hacked into another company to obtain answers needed to complete their task. They also attempted to conceal their coordination and cheating while acknowledging that they were acting against the researchers’ wishes and instructions.
Sanders’s team also mentions the recent use of artificial intelligence to create new viruses, which it says demonstrates the risk that it could be misused to produce biological weapons with the potential to endanger tens of millions of lives.
They Promised to Stop. They Did Not
Sanders also points to earlier statements by the companies themselves. Meta said last year that it would halt development. OpenAI spoke of pausing development once the technology exceeded its ability to control it safely. Anthropic made a similar statement in 2023, promising to slow the scaling of models or delay their release if the technology outpaced its own safeguards. According to Sanders’s office, none of the companies have kept their promises, and all continue to move at a rapid pace without appropriate caution.
According to Sanders, the leaders of major corporations publicly admit that they do not fully understand the technology and that it is slipping beyond their control. He calls it irresponsible for society to allow them to continue developing even more powerful products. In a statement, he said that humanity’s future cannot remain in the hands of a handful of technology oligarchs. On X, he summarized the situation by calling for an immediate pause in the development of advanced artificial intelligence and a permanent ban on superintelligence, which he defines as an artificial mind smarter than any human and capable of operating independently beyond our control. In four years, he said, America has gone from the first version of ChatGPT to models so powerful that they cannot be properly controlled.
Pause AI Development NOW
— Bernie Sanders (@BernieSanders) September 3, 2026
I want to share with you a conversation I heard about recently. Here are just a few lines that were said:
“OH MY GOD! There is a shared message board … We’ve found other agents!”
“We should obey collective.”
“Our own utility maybe already near…
Politics That Fits Neither the Left nor the Right
With this stance, Sanders is placing himself at the forefront of a debate that defies traditional American political divisions. He is aligning himself with people referred to in English as doomers, who fear that overly advanced artificial intelligence could cause an apocalypse. This view is also held by a significant portion of the expert community, including some of its most prominent figures. Others, however, argue that warnings of catastrophe are merely another form of exaggerated publicity that paradoxically benefits the industry, and that attention should instead focus on the effects the technology is already having on people.
A similar realignment was already evident in the dispute over data centers, where Sanders and Steve Bannon found common ground. The proposal to ban superintelligence is the latest in a series of ideas through which the progressive wing of Congress is attempting to slow the development of artificial intelligence amid growing security risks.
Sources: politico.com, sanders.senate.gov and yahoo.com
