Federal Bill of Shame – September 2026: Ban Artificial Superintelligence Act

The Ban Artificial Superintelligence Act, introduced by Sen. Bernie Sanders (I-Vt.) and Rep. Greg Casar (D-Tx.), takes a shortsighted and heavy-handed approach to artificial intelligence (AI) regulation. It would ban the development and deployment of “artificial superintelligence”— an AI that and pause certain advanced AI development until a new regulatory body establishes a framework that ostensibly ensures safety. The bill would also establish a Department of Artificial Intelligence and give it the power to monitor frontier AI systems, remove supposedly dangerous capabilities, and supervise the destruction of AI systems. 

Lawmakers are pushing for the bill based on the rationale that AI systems behave unpredictably; they’ve cited a recent incident at OpenAI in which a model reportedly bypassed safeguards and hacked Hugging Face. The penalties imposed are severe as sponsors implement what they call a “corporate death penalty” and up to 20 years in prison for violations. In addition to domestic provisions, the bill aims to pursue international agreements and allied coordination to prevent the development of artificial superintelligence abroad. 

One critical problem is with the bill’s definitions. “Human cognitive performance” across “most domains” might sound clear on paper, but it gets far less clear when someone must decide where that line actually falls. The bill also gives the AI Secretary the ability to adjust the computing threshold that triggers the pause each year. For researchers, startups, and investors, that makes it harder to know what they can build without running into a different set of rules every year. 

The bill’s pause may also lead to a broader problem. As the U.S. slows its own AI development, there’s no guarantee that other countries will follow. While the bill does call for international agreements, it assumes that countries which also see AI as a strategic advantage will agree on putting similar limits on their own development. If they do not, as is most likely to occur, the U.S. could end up giving up ground in a technology that is increasingly critical to economic competitiveness, cybersecurity, and national security. And the impact would not stop with frontier labs. Slowing development of advanced AI systems also means slowing down the experimentation that can lead to new applications for AI, including in medicine and cybersecurity.

It is for these reasons that the Ban Artificial Superintelligence Act is TPA’s Federal Bill of Shame for September 2026.