U.S. Bill Seeks Permanent Ban on Superintelligent AI, New Federal Agency

A proposal introduced in the U.S. Congress this week would permanently prohibit the development of artificial superintelligence and establish a new federal agency to oversee frontier AI systems, according to The Decoder. The bill was introduced on September 23, 2026, and was reported by The Decoder on September 24, 2026 — the same day the Future of Life Institute published its statement urging tighter controls on the race toward superintelligent systems.

What Happened

According to The Decoder, the bill would make it illegal to develop systems classified as artificial superintelligence and would create a dedicated federal body tasked with monitoring and enforcing that prohibition. The report does not detail how the legislation defines the threshold between advanced AI and superintelligence, a definitional problem that has dogged similar governance efforts for years. The bill was introduced by Senator Bernie Sanders and Representative Greg Casar on September 23. It remains unclear whether the measure has bipartisan backing or a scheduled committee hearing.

On the same day, Future of Life Institute published a statement in which world leaders gathered at the United Nations called for coordinated limits on the pursuit of superintelligent AI systems. The organization, which has previously campaigned for pauses on frontier model training, framed the statement as a call for governments to treat the superintelligence race as a shared governance challenge rather than a competitive advantage to be won unilaterally. The statement did not disclose a specific list of signatory nations or a proposed enforcement mechanism.

Taken together, the two developments arrived within hours of each other and point toward the same underlying anxiety: that AI capability is advancing faster than the legal and institutional frameworks meant to govern it. Neither report specifies a timeline for when a ban, an agency, or an international agreement might actually take effect.

Why It Matters

A domestic ban paired with a UN-level statement on the same day signals that superintelligence governance has moved from research-lab debate to formal political agenda in multiple capitals simultaneously. Any new federal AI agency would inevitably have to grapple with compute governance, since frontier model capability is measured largely by access to advanced processors and data center scale. That reality ties AI policy directly to the same dynamics driving semiconductor market concentration and bubble risk across the AI infrastructure buildout, where a handful of chip suppliers and cloud operators effectively determine who can attempt to build superintelligent systems in the first place. Whether lawmakers can define and enforce a ban on a category of technology that has no settled technical definition remains an open question neither report answers.

Background

Legislative interest in compute-based AI oversight follows years of policy attention on chips as strategic assets, a shift traced in MachineHours' The Global Chip Shortage entry. That crisis exposed how concentrated semiconductor manufacturing had become and pushed governments to treat production capacity, and by extension the hardware that trains large AI models, as infrastructure worth regulating rather than a purely commercial commodity.

What to Watch

Watch whether the bill receives a committee hearing or stalls without sponsor names attached, and whether the proposed federal agency's authority extends to compute thresholds rather than just software behavior. Also watch how many governments formally endorse the Future of Life Institute statement in the weeks following the UN gathering, since a symbolic statement without binding commitments carries different weight than a coordinated treaty. Any legislative language that attempts to define superintelligence in enforceable terms will likely draw scrutiny from AI labs and researchers who currently use the term loosely.