Senator Bernie Sanders and Representative Greg Casar introduced the Ban Artificial Superintelligence Act on September 23, a federal bill seeking to prohibit the development or deployment of systems it defines as superhuman artificial intelligence, temporarily halt some development of advanced systems and establish a new federal body to oversee the field.
This is a bill introduced in Congress, not a law that has taken effect. As of publication, it imposes no new restrictions on companies or developers, and no new government department has been established under it.
What the proposal seeks to do
According to the bill's text and its sponsors' announcements, the proposed law would prohibit developing or deploying systems meeting its definition of artificial superintelligence. The definition refers to systems exceeding human ability in most cognitive domains, or systems capable, under the law's criteria, of posing an extreme threat to humanity's ability to control them.
The proposal also includes a pause on developing certain advanced AI systems until a federal regulatory framework is established with testing, oversight and approval procedures.
Another component is establishing a cabinet-level Department of Artificial Intelligence. The proposed body would oversee frontier systems throughout their lifecycle, identify capabilities defined as dangerous and supervise the removal or destruction of systems violating the ban.
The proposed penalties are unusually severe
The bill includes substantial sanctions. According to the sponsors' announcement, corporations violating the prohibitions could face measures up to corporate dissolution, and individuals could face prison sentences of up to 20 years.
These penalties do not currently apply. They are part of a proposal that still needs to pass through the congressional legislative process, including committee discussions, votes and possible changes to the text.
Why the proposal is broader than a typical AI law
Most US regulatory discussions focus on particular uses, transparency obligations, discrimination risks, security or model providers' liability. Sanders and Casar's proposal goes further, trying to draw a red line around the development itself of a future category of systems.
The sponsors argue that superhuman systems could create extreme risks and therefore require a framework preventing their development in advance. This is a political and policy position expressed by the bill's sponsors. It is not an agreed legal or scientific determination that existing systems already meet the definition of superintelligence.
It is unclear how thresholds would be implemented in practice
One central question will be how to legally define a system exceeding human ability "in most domains," and how to measure this threshold in a way that enables enforcement.
The term advanced AI development in the context of the pause also requires measurement and standardization mechanisms. The proposal describes a future regulatory framework, but practical implementation would require detailed tests, evaluation procedures and powers.
It is also unclear how a new body would coordinate its activities with federal agencies already dealing with cybersecurity, standards, competition, consumer protection and national security.
Current status: a proposal, not binding policy
Sanders and Casar announced their intention to advance the initiative earlier in September, but on September 23 they announced the introduction of the bill itself. This stage is important in parliamentary terms, but remains far from final legislation.
No binding schedule for committee discussion has been published, and the level of support in the House or Senate is uncertain. There is also no basis at this stage to determine whether the proposal will remain in its current form, be amended or fail to advance.
The proposal's immediate significance is therefore mainly political: it brings into Congress a regulatory model seeking not only to oversee advanced AI but to ban an entire category of future systems. If it advances, the debate is likely to focus on definitions, oversight powers and how to balance research, competition and potential risks.