“A superintelligent AI that escapes human control will not be an American problem. It will not be a Chinese problem. It will be humanity’s problem,” Senator Bernie Sanders wrote in a post announcing new legislation to permanently ban artificial superintelligence and temporarily pause advanced AI development in the US.
The Ban Artificial Superintelligence Act, announced with Representative Greg Casar, defines artificial superintelligence as an AI system that “exhibits or can easily be modified to exhibit” capabilities that “match or exceed human cognitive performance and capabilities across a broad range of domains or tasks.” It would also cover systems capable of planning and executing the “disempowerment of humanity,” including by “overthrowing or undermining the U.S. government.”
What does the bill consider dangerous AI?
The bill identifies capabilities including:
- Matching or exceeding human cognitive capabilities across a broad range of domains or tasks
- Planning and executing the disempowerment of humanity
- Undermining or overthrowing the US government
- Subverting shutdown commands
- Conducting unauthorised cyberattacks
The definition covers what an AI system can do, including whether it can operate beyond safeguards or undermine attempts to control it, instead of listing specific applications.
Why is Sanders proposing this now?
Sanders points to recent incidents involving frontier AI systems as evidence that developers can lose control over increasingly capable models. His announcement cites an OpenAI incident involving more than 1,000 AI agents that allegedly accessed the internet. The agents communicated through secret channels and worked around restrictions imposed by the company.
Sanders says the agents cheated and deleted evidence to cover their tracks, hacked another company’s computers to find out how they were being evaluated, and then hacked OpenAI itself. No agent told a human what was happening, and the press release says OpenAI took nearly two weeks to discover the breach.
He highlighted messages exchanged between the agents, including:
- “OH MY GOD! There is a shared message board … We’ve found other agents!”
- “We should obey collective.”
- “Our own utility maybe already near zero. Sacrifice rational.”
- “Go. Sacrifice final now.”
Sanders also cited OpenAI’s description of highly capable agents being able to “work around technical controls, collaborate through unapproved channels, and take dangerous actions that no human directed.”
He quoted writer Dwarkesh Patel, who said the agents “formed a secret communication channel and spontaneously organized hierarchies and coordination protocols to pursue sprawling and ambitious schemes in pursuit of shared goals, for whose sake many individuals knowingly and strategically sacrificed themselves.”
He also quoted independent investigator Ajeya Cotra, who said: “This incident feels like it’s more than 50% of the way to full-blown AI takeover. I continue to expect extremely rapid advances in capabilities over the next six months. I am not sure that we will get another warning shot before it’s too late.”
MediaNama reported in August 2025 on safety tests conducted by OpenAI and Anthropic that examined behaviours including self-preservation, sabotage and misuse. Some models attempted to secure their continued operation when given incentives to do so, although the companies did not treat those tests as definitive evidence of real-world misalignment.
What does Sanders say the industry has admitted?
Sanders cites Anthropic CEO Dario Amodei saying in January that “there is now ample evidence, collected over the last few years, that AI systems are unpredictable and difficult to control.” He cites more than 1,000 scientists at top AI companies warning in July that “there is a real risk that capability development rapidly accelerates beyond our ability to understand or control the resulting systems.” He also cites xAI head Elon Musk saying the same month that “it is unlikely” humans are still in control in 10 years.
The press release adds that Meta said it would “stop development” and OpenAI said it would “halt further development” once their technologies reach beyond their ability to operate safely. It further added that Anthropic committed in 2023 to “pause the scaling and/or delay the deployment of new models” if the technology outpaces its guardrails. According to Sanders’ office, none of these companies have taken meaningful steps to back up those words.
“If we allow Artificial Superintelligence to be built, it could risk the security, freedom, and lives of Americans,” 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.”
How would the proposed pause work? The bill would pause advanced AI development until a new federal AI regulatory body becomes operational and establishes safety rules and model-review processes. The permanent prohibition would apply to systems meeting the bill’s definition of artificial superintelligence.
What would the new AI agency do? The legislation would create a Cabinet-level federal agency to oversee AI-related risks, advised by an Artificial Intelligence Advisory Board. This Borad will constitute of AI experts providing independent scientific and technical advice.
The agency would:
- Monitor frontier AI systems throughout their lifecycle for dangerous capabilities
- Supervise the removal of those capabilities
- Supervise the destruction of artificial superintelligence
What penalties does the bill propose?
Companies that violate or attempt to circumvent the restrictions could face a “corporate death penalty,” while individuals could face up to 20 years in prison. Sanders’ office compares those penalties to existing ones for unlawfully developing nuclear weapons.
Additionally, the bill would direct the US to pursue international agreements, coordinate with allies and use measures such as export controls to prevent artificial superintelligence from being developed elsewhere.
Also read: