Senator Bernie Sanders, an Independent from Vermont, and Representative Greg Casar, a Texas Democrat who chairs the Congressional Progressive Caucus, introduced the Ban Artificial Superintelligence Act on September 23, 2026. The proposal represents a sweeping attempt to draw a legal boundary around how far artificial intelligence development should be allowed to go in the United States.
At its core, the bill separates artificial superintelligence from other advanced AI. It would permanently prohibit the development and deployment of artificial superintelligence while placing a temporary pause on development of advanced AI systems until federal safety rules are established. Once those rules exist, advanced AI systems could not be deployed without federal approval.
What Is the Ban Artificial Superintelligence Act?
The proposal therefore combines a permanent ban at one end of the capability spectrum with a regulatory system for advanced AI that falls short of the bill's superintelligence threshold.
Sanders framed the proposal as a response to the possibility that increasingly capable systems could eventually operate beyond meaningful human control. In comments quoted by the Associated Press, he asked: "Do we really want to develop a super intelligence that when it becomes smarter than human beings could act independently of human control?"
Fortune also quoted Sanders saying, "It doesn't take a genius to say, slow it down."
How the Bill Defines Artificial Superintelligence
One of the most important parts of the legislation is its definition of artificial superintelligence, because that definition determines what technology would face a permanent prohibition rather than temporary restrictions.
The bill defines artificial superintelligence as a system that exceeds human cognitive ability, or a system capable enough to plan and carry out the destruction or disempowerment of humanity.
That language makes the bill fundamentally about capability rather than a particular company, model family or technical architecture. The prohibition is aimed at systems reaching the specified threshold, rather than artificial intelligence as a whole.
The definition also distinguishes the proposal from a blanket prohibition on advanced AI. Systems that are advanced but do not meet the superintelligence definition would instead fall under the bill's temporary pause and approval requirements.
This distinction matters because the legislation does not treat every powerful AI system as artificial superintelligence. Its structure anticipates a category of advanced systems that could continue to exist under federal safety rules and approval, alongside a separate category that lawmakers sponsoring the bill want prohibited entirely.
That creates two different regulatory questions: whether an advanced system is safe enough for federal approval, and whether a system has crossed the bill's threshold for artificial superintelligence.
How Federal Approval for Advanced AI Would Work
The bill would temporarily pause development of advanced AI systems until a federal regulator has established safety rules. After that regulatory framework is in place, an advanced AI system would need federal approval before deployment.
To administer that system, the legislation would create a new Department of Artificial Intelligence. The department would be responsible for writing and enforcing the required rules.
The proposal therefore goes beyond asking AI developers to adopt voluntary safeguards. It would establish a federal gate between development and deployment for systems covered by its advanced-AI provisions.
The bill also attaches criminal consequences to violations. According to the legislation described in the Associated Press coverage, violations could carry penalties of up to 20 years in prison.
That enforcement structure is one of the clearest signs of how differently the proposal treats advanced AI from an ordinary technology product. Development could be paused while rules are written, deployment could require affirmative government approval, and breaking the resulting framework could expose violators to criminal penalties.
The bill would also make the regulatory question an ongoing federal responsibility rather than a one-time legislative decision, because the proposed department would write and enforce the safety rules governing advanced systems.
Why Sanders Wants an International AI Agreement
The legislation does not stop at domestic regulation. It directs the United States to pursue international agreements intended to prevent artificial superintelligence from being developed anywhere in the world.
Sanders compared that idea with Cold War nuclear arms control. The comparison reflects a central practical challenge within the proposal itself: a prohibition applying only inside the United States would not prevent developers elsewhere from pursuing systems that meet the bill's definition of artificial superintelligence.
The international element therefore extends the bill's objective beyond controlling where or how US companies build AI. Its stated goal is a wider prohibition on the development of superintelligence.
That is a substantially different objective from regulating the deployment of advanced systems in the American market. Domestic approval rules could be enforced through a US regulator, while preventing superintelligence worldwide would depend on international agreements.
The bill consequently links two forms of governance: a national regulatory system for advanced AI and international coordination around the technology the proposal seeks to prohibit permanently.
Sanders' arms-control analogy also shows how the legislation's sponsors are framing superintelligence: not simply as another commercial technology requiring ordinary oversight, but as a capability whose risks could require agreements between governments.
AI Workers Supporting the Superintelligence Ban
Several employees of leading artificial intelligence companies have endorsed the proposed legislation, according to the Associated Press coverage. Their support adds an industry perspective to a measure that would impose significant restrictions on advanced AI development.
Juan Felipe Ceron Uribe, a researcher on OpenAI's Safety Systems team, said superintelligence could "either go extremely right or extremely wrong." The statement reflects the wide range of outcomes supporters of the bill associate with highly capable systems.
Swante Scholz, a software engineer at Google DeepMind, offered a more severe assessment. He said the "most likely outcome of the current trajectory is an existential catastrophe for humanity."
Those comments are arguments from individual employees supporting the proposal, rather than findings established by the bill itself. They nevertheless illustrate why some people working inside prominent AI organizations are backing restrictions that could directly limit development of increasingly capable systems.
The bill arrives during a period of intense discussion about where advanced AI could lead and how early governments should intervene. Its supporters are not merely calling for stronger testing or reporting rules. They are supporting a framework that would draw a permanent legal line at artificial superintelligence while requiring government oversight for advanced systems below that threshold.
What the AI Bill Does Not Do and Arguments Against It
The Ban Artificial Superintelligence Act does not prohibit all artificial intelligence. Its permanent ban applies to artificial superintelligence as defined by the legislation, while advanced AI below that threshold would be subject to a temporary development pause followed by federal safety rules and approval requirements.
It also does not establish that artificial superintelligence currently exists. Instead, the proposal defines the capability it seeks to prohibit and creates restrictions intended to apply as AI systems advance.
The legislation's political prospects are also uncertain. The Associated Press reported that the bill faces long odds in a Republican-controlled Congress, where lawmakers have struggled to agree on AI rules that are far less sweeping. Democrats also do not have a settled position on the issue. For the wider congressional picture beyond the technology beat, our sister title The NE Times follows US politics and world affairs with independent commentary and analysis.
Those circumstances form an important part of the case against expecting the proposal to become law in its introduced form. The bill goes further than narrower approaches by proposing a permanent prohibition, a new federal department, pre-deployment approval and criminal penalties.
Opposition can therefore center on the breadth of the intervention itself, even as supporters argue that unusually capable AI requires unusually strong safeguards. The congressional difficulty described by the Associated Press means the immediate debate is not simply about how the rules would work, but whether lawmakers can agree that restrictions of this scale should exist at all.
Why the Superintelligence Bill Is Getting Attention Now
The bill emerged during a particularly active week for discussion around advanced AI. On September 23, 2026, it trended on X alongside broader "Super Intelligence" discussion.
Its introduction also came one day after Anthropic released Claude Opus 5.5, a cheaper frontier model. The timing creates a sharp news sequence: a major capability release followed the next day by legislation seeking to prohibit artificial superintelligence and tightly regulate advanced AI. There is no established causal link between those two events.
Public concern about AI is also extending beyond theoretical future systems. Associated Press polling cited in the coverage found that 53 percent of Americans were extremely or very concerned about AI's environmental impacts, compared with 41 percent a year earlier.
That polling does not measure support for the Sanders-Casar bill, and concern about environmental impacts is distinct from concern about superintelligence. It does, however, show that debates over AI policy are taking place while public attention extends to the technology's broader effects.
The legislation now puts a particularly expansive policy option into that debate. Rather than focusing only on individual applications of AI, it asks Congress to decide whether one potential level of machine capability should be prohibited permanently, whether advanced systems should require federal permission before deployment, and whether the United States should pursue international agreements aimed at making the prohibition global.
FAQ
Who introduced the Ban Artificial Superintelligence Act?
Senator Bernie Sanders, an Independent from Vermont, and Representative Greg Casar, a Texas Democrat and chair of the Congressional Progressive Caucus, introduced the Ban Artificial Superintelligence Act on September 23, 2026. The legislation would permanently prohibit artificial superintelligence while creating a separate regulatory framework for advanced AI systems that remain below the bill's superintelligence definition.
Would the bill ban all artificial intelligence?
No. The permanent prohibition applies specifically to artificial superintelligence as defined in the bill. Advanced AI systems below that threshold would face a temporary development pause until federal safety rules are established. After those rules exist, covered advanced AI systems would require federal approval before deployment, rather than being permanently prohibited.
How does the bill define artificial superintelligence?
The legislation defines artificial superintelligence as a system that exceeds human cognitive ability, or one with sufficient capability to plan and carry out the destruction or disempowerment of humanity. That definition determines which systems would face the proposed permanent prohibition rather than the temporary restrictions and federal approval process applying to advanced AI.
What would the Department of Artificial Intelligence do?
The bill would establish a new Department of Artificial Intelligence responsible for writing and enforcing federal safety rules for advanced AI. Development of advanced systems would be temporarily paused until those rules were established. Once the framework existed, covered advanced AI systems would need federal approval before they could be deployed.
What penalties does the AI bill propose?
The legislation provides for criminal penalties of up to 20 years in prison for violations, according to the Associated Press coverage of the proposal. That penalty sits alongside the bill's broader enforcement framework, which includes a permanent superintelligence prohibition, temporary restrictions on advanced AI development and federal approval before advanced systems can be deployed.
Is the Ban Artificial Superintelligence Act likely to pass Congress?
Its prospects are uncertain. The Associated Press reported that the bill faces long odds in the Republican-controlled Congress, where lawmakers have struggled to reach agreement on AI regulations that are considerably less sweeping. Democrats also do not have a settled position. Those political conditions make the bill part of an unsettled congressional debate over how extensively advanced AI should be regulated.
