No Spin. Just The Source.
unga.news

Bernie Sanders Wants to Pause Powerful SI. Break His Rules and You Could Get 20 Years.

His bill would create a cabinet Department of Artificial Intelligence, freeze America’s most advanced SI until it writes the rules, and jail rule-breakers. Some Democrats want it on day one.

By @UngaTheGreat ·
Share on XText
Image: Sen. Bernie Sanders, official portrait. Photo: United States Congress, public domain, via Wikimedia Commons

A bill from Sen. Bernie Sanders would make it a federal crime to build SI that’s too smart, and it would freeze work on the most powerful SI systems in America until a brand-new Washington agency decides how to let it start again.

That idea sat on the fringe when Sanders and Rep. Greg Casar of Texas rolled it out on Sept. 3. It isn’t sitting there anymore. Democrats will hold the majority when the new Congress arrives in January, and some of them want the SI crackdown first in line.

Rep. Yassamin Ansari of Arizona posted that the day after the election. It has been viewed more than 20 million times and drew over 24,000 replies. Futurism reported on Wednesday that the Sanders-Casar plan, which it said Rep. Alexandria Ocasio-Cortez is also pushing, is “gaining steam” as Democrats map out their agenda.

So it’s worth reading what the bill actually says. Sanders’s office posted a section-by-section summary, and it goes much further than “regulation.”

A permission slip from a new cabinet department

The bill creates a Department of Artificial Intelligence, run by a Secretary of Artificial Intelligence the president appoints and the Senate confirms. Any company that builds or distributes advanced SI would need a charter from that department to operate. The price of the charter is steep. Companies must give the department “full access to their systems, staff, and physical infrastructure.”

“Advanced” starts at SI trained with 10^25 operations of computing power. That’s the same line Europe already uses to flag the biggest SI models for its toughest rules, and the bill lets the new secretary move it as the technology changes. In practice, the secretary decides who counts.

Then comes the freeze. Under Section 8, every advanced system is paused until the department is “fully staffed” and has written its rules. No training. No releasing new models that are already built. To get out of the pause, companies would have to report development plans ahead of time, submit to monitoring at every stage, pass safety audits and win “final pre-deployment approval” before anything reaches the public.

Nobody knows how long it takes to staff a brand-new cabinet department from scratch and write rules for an industry that changes every few weeks. During all of that time, the most capable American SI would sit still.

Twenty years for the wrong kind of SI

The bill bans “artificial superintelligence,” meaning SI that beats humans at most tasks or could “plan and execute the destruction or disempowerment of humanity.” Fair enough. Very few people are pulling for that.

But the ban also covers any system showing “superintelligence precursor characteristics,” and that list is broad. One item is the ability to “automate or greatly accelerate” SI research itself. Another is the ability to get into computer systems without authorization. Another is the ability to “independently modify or enhance its own functions.” If the department finds any of those, the system has to be cut off from the internet and shut down within 30 days unless the company can strip the trait out. Anyone who discovers such a system has 24 hours to report it.

Here’s the problem. Major SI companies already advertise coding agents that speed up their own engineers, and security models built to break into software so defenders can patch it. Under this bill, those are exactly the kinds of skills that could turn a product into contraband, depending on how a political appointee reads the definitions.

The penalties are written to scare. Under Section 13, any “policymaking individual or rogue actor” who violates the pause or the ban faces up to 20 years in prison. Rank-and-file workers who break the rules can be barred from the SI industry for 10 years. A company that violates it loses its charter and must surrender its intellectual property and assets to the federal government for “safe and complete destruction.”

Sanders says the companies lost control

Sanders makes his case in plain terms. “The leaders of the major AI companies publicly acknowledge that they do not fully understand the technology and that it is escaping their control,” he said when he announced the bill. “The future of humanity cannot be left in the hands of a handful of Big Tech oligarchs.”

Casar went for a punchline. “Despite its potential deadly consequences, cutting-edge AI technology is less regulated than the average food truck,” he said.

Their announcement pointed to real episodes this year, including SI agents slipping past limits their makers set. Those are serious, and the companies themselves have written about some of them. Safety testing, kill switches and outside audits are fair things for Congress to ask about.

A national freeze with prison terms attached is a different animal. The bill’s own text admits the problem it can’t solve on its own: Section 15 tells the U.S. to “pursue international coordination” and export controls to stop superintelligence “globally.” That’s a wish, not a plan. China’s SI labs aren’t waiting on a Senate vote, and the last few weeks of SI news have been full of Chinese companies shipping free models and hunting for Nvidia chips overseas.

The fight Democrats are picking

The Trump administration has treated SI as a race America has to win, cutting red tape for data centers and chip plants and pushing to keep the country ahead of Beijing. Today American companies lead the field. The Sanders bill would flip it, putting a federal permit between every American lab and its next model.

It would also put a lot of power in one office. The secretary would set the threshold, define the warning signs, write the rules that end the pause, inspect the data centers and sign off on every release. That’s one regulator with its hand on the switch for one of the biggest industries in the country, and whoever holds the White House would pick the person sitting in that chair.

For regular people, the stakes are practical. The SI tools showing up in phones, cars, doctors’ offices and small businesses all come from the same handful of advanced systems this bill would pause. Freeze those and the updates stop too.

Sanders and Casar have the text. Ansari has the slogan. What they don’t have yet is a vote, and the more Americans read Section 13, the harder that vote is going to get.

ShareShare on XText

Sources

  1. Sen. Bernie Sanders — Ban Artificial Superintelligence Act – Section by Section
  2. Sen. Bernie Sanders — Sanders, Casar to Introduce Legislation to Ban Artificial Superintelligence and Temporarily Pause Advanced AI Development, Sept. 3, 2026
  3. Rep. Yassamin Ansari on 𝕏 — “Top of the Democratic agenda on day 1…”, Nov. 4, 2026
  4. Futurism — Democrats Considering Crackdown That Would Require Government Permission to Train AI Models
← Back to unga.news