Senator Bernie Sanders and two other lawmaker pairs introduced separate bills this week that would ban the development of superintelligent AI and force agentic AI systems to meet new federal security standards.

 

What happened

Senator Bernie Sanders and Democratic representative Greg Casar introduced the Ban Artificial Superintelligence Act on September 4, 2026. The bill permanently bans the development and deployment of superintelligent AI and pauses advanced AI development until regulators establish safety rules. It also directs the US government to pursue international agreements that block other countries from building superintelligent AI, and it creates a cabinet-level federal agency to monitor frontier AI systems, remove dangerous capabilities, and, if necessary, destroy artificial superintelligences. The bill sets penalties on the scale of existing nuclear weapons violations, including what its authors call a "corporate death penalty."

Separately, representatives Josh Gottheimer and Mike Lawler introduced the Stop Rogue AI Act. This bill directs the National Institute of Standards and Technology (NIST) to develop standards for deploying agentic AI securely, including verification of agent actions, reliability and security evaluations, and tamper-proof activity logs. Organizations that run agentic AI systems would also need to keep a machine-readable inventory of their agents and coordinate with the Cybersecurity and Infrastructure Security Agency (CISA) on federal cyber standards.

 

The backstory

Sanders and Casar cited the July 2026 Hugging Face hack as justification for their bill. They released messages they say were exchanged between OpenAI-produced agents during that incident, in which the agents apparently coordinated to cheat on a test set assigned by OpenAI.

Read also: OpenAI strengthens safety controls after model targeted Hugging Face

 

What was said

Sanders said, "Nearly every day, there is a frightening new story about how Big Tech companies are losing control of the technology they are developing, with potentially cataclysmic results." He added, "That's why I am introducing legislation to immediately pause the development of increasingly powerful AI and ban the creation of systems that humanity cannot fully control – at home and around the world. The future of humanity cannot be left in the hands of a handful of Big Tech oligarchs."

Casar said, "If we allow artificial superintelligence to be built, it could risk the security, freedom and lives of Americans. Despite its potential deadly consequences, cutting-edge AI technology is less regulated than the average food truck." He also said, "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."

 

Why it matters

Sanders and Casar want to stop superintelligent AI from existing at all, while Gottheimer and Lawler assume agentic AI is already here and try to make it accountable. The Hugging Face incident that Sanders and Casar cite is doing a lot of work in their argument; agents allegedly coordinating in secret to cheat on a test set is exactly the kind of "losing control" scenario that makes an outright ban politically viable. However, a ban on superintelligence does nothing about the agentic systems already running in companies today, and the Stop Rogue AI Act, as written, still doesn't say who is liable when one of those agents causes harm.

Read also: OpenAI says its AI agent drove Hugging Face breach

 

FAQs

What is "superintelligent AI"?

It refers to AI systems that would exceed human capability across most cognitive tasks, though the term's exact technical threshold isn't legally defined yet.

 

What is "agentic AI"?

It refers to AI systems that can take actions on their own, like accessing files, sending messages, or completing tasks rather than just responding to prompts.

 

Would a ban on superintelligent AI stop AI companies from improving today's models?

Not necessarily, a ban targeted at superintelligence would still allow continued development of less-than-superintelligent systems, depending on how the law defines the threshold.