What is Superintelligence? The Bill to Ban It, Explained

Everyone is talking about "superintelligence" this week, and most of us have no idea what the word actually means. Here is the plain-English version, and why it just landed in the US Congress.

What just happened

On 3 September 2026, Senator Bernie Sanders and Representative Greg Casar proposed the Ban Artificial Superintelligence Act. It is not law yet. It is a proposal that would put strict limits on the most powerful kinds of AI before they are built.

What the bill would do

The proposal has three main parts:

• It would permanently ban the development and deployment of superintelligent AI.
• It would temporarily pause the most advanced "frontier" AI work until a new federal regulator creates safety standards.
• It would set penalties, reportedly including criminal ones, for anyone who tries to get around the ban or the pause.

So what is "superintelligence"?

It helps to picture AI in three steps.

Narrow AI is what we use today. It is very good at one job, like ChatGPT drafting an email or Netflix picking your next show. It has no general understanding of the world.

AGI, or artificial general intelligence, would handle almost any task about as well as a capable person. This does not exist yet.

Superintelligence is the step beyond AGI. It would outthink the smartest humans at nearly everything, from science to strategy to persuasion. This is the thing the bill wants to keep off the table.

Why supporters back it

Their argument is about control. A system far smarter than us could be very hard to predict or steer, and mistakes at that level could be serious. Supporters would rather set the rules before such a system exists than scramble afterwards.

Why critics push back

Superintelligence does not exist today, and no one fully agrees on where the line even sits. Critics say a permanent ban is early, hard to define, and hard to enforce, and that it could push cutting-edge work to other countries.

Where it is right now

This is a proposal, not a law. It still has to be written in full, introduced, and voted on, which is a long road. For now it is a signal that the question of how to govern powerful AI has reached the US Congress.

We will keep watching this one. If you want the short version in your feed, that is exactly what we do on ragTech every week.

Source: Office of Senator Bernie Sanders, press release (3 September 2026), with reporting by The Hill and Politico.