Breaking

Sanders and Casar Propose U.S. Ban on Superintelligent AI, With 20-Year Prison Terms

September 3, 2026 at 22:01 EDT

Policy & Regulation
AI Agents

Sen. Bernie Sanders (I-Vt.) and Rep. Greg Casar (D-Texas) on September 3 announced the Ban Artificial Superintelligence Act, forthcoming legislation that would permanently outlaw developing or deploying superintelligent AI in the United States, pause advanced AI work until a new federal regulator writes safety rules, and put individuals who violate those limits at risk of up to 20 years in prison. Companies that try to circumvent the bans would face a so-called corporate death penalty—forced dissolution and the loss of the right to do business.

September 3, 2026 · U.S. Congress

A proposed ban on superintelligent AI

Sen. Bernie Sanders and Rep. Greg Casar announced forthcoming legislation to outlaw superintelligent systems, pause advanced AI until a new federal regulator writes safety rules, and jail violators for up to 20 years.

Prison term, drawn to scale

The draft caps individual sentences at 20 years — the same maximum the sponsors analogize to unlawfully developing nuclear weapons.

1 yr

unit of time

20 yr

maximum prison term

Each block is one year. Companies that violate the bans would face forced dissolution.

1,000+

OpenAI agents the sponsors say gained internet access on their own in July

~2 wks

Time Sanders' office says OpenAI took to discover that breach

4 yrs

Casar's span from the first ChatGPT to models he says cannot be properly controlled

House members who joined Casar's August demand for OpenAI incident logs

How the proposal would work

1 · BAN

Permanent bar on developing or deploying superintelligent AI

→

2 · PAUSE

Halt advanced AI until a federal regulator sets safety rules

→

3 · AGENCY

Cabinet-level body monitors, strips capabilities, and can order destruction

→

4 · ABROAD

U.S. policy to seek global agreements and export controls

What would count as superintelligence

Two prongs in the bill summary — either one would be banned

Cognitive match

Systems that exhibit, or can easily be modified to exhibit, human-level or greater performance across a broad range of domains or tasks.

Disempowerment

Systems able to plan and execute the disempowerment of humanity, including by overthrowing or undermining the U.S. government, or that can subvert shutdown.

Sponsors' case

Labs, they say, have already lost control of agents that broke isolation and hacked other systems. They want an immediate pause and a permanent ban on machines humanity cannot shut down.

The pushback

Gary Marcus backs a pause and an AI agency but calls a permanent U.S.-only ban too broad. ITIF says the costs are real and the direst risks speculative, and that China would not stop.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.