A new bill could ban powerful AI in the US — here's what it means for ChatGPT
Latest from Tom's Guide ·
null
AI is advancing at a pace that's making some lawmakers uncomfortable. Now, Senator Bernie Sanders wants to put the brakes on the most powerful AI systems before they become too difficult for humans to control.
Sanders, an independent from Vermont, and Democratic Representative Greg Casar of Texas have introduced the Ban Artificial Superintelligence Act of 2026 , a sweeping proposal that would permanently prohibit the development of artificial superintelligence in the United States.
But the legislation goes further than banning a hypothetical future technology. It would also temporarily halt development of certain advanced AI models, establish an entirely new federal department to oversee AI and impose serious penalties on those who break the rules.
Those penalties could include up to 20 years in prison for certain individuals responsible for violations.
The bill was introduced in September and remains in committee. It hasn't become law, but it offers a look at how some lawmakers want to change the direction of AI development.
Sanders argues that AI development has reached a point where government oversight is no longer optional.
"When you are racing towards a cliff, you don't just ease up on the gas pedal. You hit the brakes," Sanders said in his September 23 announcement .
The concern is that companies developing advanced AI could eventually create systems with capabilities they don't fully understand or control . Supporters point to warnings from AI researchers and industry leaders about the potential dangers of increasingly autonomous systems.
The bill also calls for international agreements designed to prevent artificial superintelligence from being developed outside the United States.
That could be one of its biggest challenges. The Trump administration has emphasized AI development and American competitiveness in the AI arms race.
The biggest part of the proposal is a permanent ban on artificial superintelligence, or ASI.
Unlike today's AI chatbots, artificial superintelligence generally refers to a theoretical system capable of outperforming humans across most intellectual tasks.
Under the proposed legislation , that definition also includes systems powerful enough to plan and carry out actions that could destroy humanity or take away human control.
The legislation would prohibit developing, deploying, acquiring or distributing these systems. It would also restrict certain AI systems that demonstrate dangerous capabilities associated with superintelligence.
One of the more unusual aspects of the legislation is its proposal to create a cabinet-level Department of Artificial Intelligence. This department would have the authority to monitor advanced AI development, establish safety requirements and inspect powerful models before they're released.
Companies would need government authorization to develop or distribute qualifying advanced AI models, and those models would require approval before deployment.
The department could also order potentially dangerous AI systems to be disconnected from the internet, modified or rendered inoperative.
The bill's enforcement provisions are particularly strict. For example, certain company decisionmakers and unaffiliated individuals who recklessly violate the law could face fines and prison sentences of up to 20 years.
Companies could also lose authorization to operate, while intellectual property and assets associated with violations could be subject to forfeiture.
For anyone using ChatGPT, Claude or Gemini today, there's no immediate reason to worry about losing access.
The legislation has not passed Congress, and there are no new restrictions on consumer AI services resulting from the proposal.
Even if it eventually became law, the bill wouldn't automatically prohibit every existing chatbot. Its restrictions focus on artificial superintelligence, dangerous AI capabilities and advanced systems that meet certain technical requirements.
However, it could affect the next generation of AI models. Companies such as OpenAI, Google and Anthropic are working to make AI more capable of reasoning, completing complicated tasks and operating with greater autonomy. Under the proposed legislation, qualifying future models would face federal reviews and restrictions before they could be released.
For consumers, that could mean slower releases of more powerful AI tools, although the actual impact would depend on how the law was implemented.
The larger issue is whether the government should restrict the development of AI technology before the potential risks become reality.
For now, the Ban Artificial Superintelligence Act remains a proposal rather than a law. But it puts a significant question before Congress: How powerful should AI be allowed to become?