Sanders Proposes Ban on Advanced AI, Including Superintelligence

Senator Bernie Sanders wants Congress to put much tighter limits on the development of advanced AI, including a permanent ban on artificial superintelligence. Sen. Sanders and Rep. Greg Casar announced the Ban Artificial Superintelligence Act on September 3, 2026. In a follow-up statement, Sanders said that violators could face up to 20 years in prison.

The proposal goes well beyond rules for AI-generated content or the specific use of chatbots. This proposed legislation could temporarily halt the development of advanced AI while new federal safeguards are created. Then, it could permanently prohibit systems that reach the level of artificial superintelligence.

The proposal also includes creating a new federal agency with the authority to monitor frontier AI and remove any dangerous capabilities that could cause future harm.

Sanders wants to put limits on how far advanced AI systems could go

The proposed legislation would give the federal government a more direct role in controlling the development of more powerful AI systems. The proposed agency would monitor frontier models throughout their lifecycle and oversee the removal of any dangerous capabilities.

For certain superintelligent AI systems found to be in direct violation, the proposed legislation also includes provisions to supervise their destruction.

There is also a significant penalty aimed at companies that violate the restrictions, including dissolving violating companies and imposing up to 20 years in prison for individuals. Sanders compared the proposed sentence with existing penalties for illegally developing nuclear weapons.

The proposal also includes calls for the US to work with other countries to prevent superintelligence from being developed abroad. This includes creating the framework for international agreements and export controls. However, this would only come into effect if the proposed legislation is passed in Congress.

The push comes amid the recent safety issues involving new AI systems

Sen. Sanders and Rep. Casar specifically mention the recent incidents involving advanced AI systems escaping or circumventing their isolated testing environments. Anthropic and OpenAI have reported cases where their AI systems gained unauthorized access to real systems during cybersecurity testing.

In the announcement for the proposed legislation, Rep. Casar said, "Despite its potential deadly consequences, cutting-edge AI technology is less regulated than the average food truck. That must change. In just four years, we have gone from the first version of ChatGPT to AI models so powerful they cannot be properly controlled. Congress should immediately ban AI systems too powerful to control."

As AI agents gain the ability to use tools, browse the internet, and carry out multi-step tasks, prior testing of whether they will stay within their intended limits has become increasingly important.

The industry is now openly debating how quickly these capabilities should advance. Anthropic CEO Dario Amodei has recently called for frontier AI companies to slow development. OpenAI CEO Sam Altman and xAI CEO Elon Musk have publicly supported his call.

Congress faces a wider debate over how much control AI companies should have

Sen. Sanders' proposed bill is only a part of a broader shift. In July, Reps. Ted Lieu and Nathaniel Moran introduced the bipartisan AI Kill Switch Act. This proposed bill would require developers of powerful AI systems to build in shutdown capabilities in the event the AI systems pose a serious risk. The two proposals seem to address different stages of the same problem.

While the AI Kill Switch Act is focused on building safeguards to suspend or shut down dangerous AI systems even after deployment, the Ban Artificial Superintelligence Act proposes federal oversight of the removal of dangerous capabilities and the subsequent destruction of prohibited artificial superintelligence systems.

Both bills are only proposals for now and are not currently law. Together, however, these proposals show that the congressional debate is moving beyond general AI safeguards and is now actively questioning whether the government should have the ability to directly intervene when advanced AI systems become too dangerous to ignore.

Affiliate links on Cult of Mac may earn us and our partners a commission.