In Historic Move, U.S. Lawmakers Introduce Legislation to Ban Artificial Superintelligence

Two progressive lawmakers have introduced an unprecedented legislative bid to draw a hard line on advanced machine learning. Sen. Bernie Sanders (I-Vt.) and Rep. Greg Casar (D-Texas) have unveiled the Ban Artificial Superintelligence Act, a sweeping draft measure designed to permanently outlaw the creation and deployment of artificial superintelligence (ASI) and temporarily freeze frontier AI development across the United States. Under the bill, violators could face up to 20 years in federal prison and corporations could face dissolution under a statutory "corporate death penalty," positioning the potential risks of unchecked AI alongside the proliferation of non-conventional weapons.

Background

To understand the significance of this move, it helps to distinguish the current state of technology from the future systems targeted by lawmakers. Today’s commercial software—from large language models to autonomous coding assistants—is broadly classified as narrow or generative AI. While proficient at processing human language, analyzing imagery, and writing code, these models remain specialized tools rather than autonomous cognitive agents.

Artificial General Intelligence (AGI), long described as the industry’s holy grail, refers to hypothetical software capable of performing any intellectual task at human-level proficiency across a broad spectrum of disciplines. Artificial Superintelligence (ASI) goes a step further: it describes a system that vastly exceeds all human cognitive capabilities, possessing the ability to self-improve iteratively, solve complex scientific problems at impossible speeds, and potentially circumvent technical guardrails.

Over the past several years, leading frontier research labs including OpenAI, Anthropic, and Meta published voluntary safety commitments, pledging to halt model scaling if developments began outstripping human capacity to evaluate or control them safely. However, critics in civil society and government have argued that voluntary pledges by private entities are structurally insufficient in a hyper-competitive race for technological dominance, leaving society exposed to catastrophic misalignment risks.

What Happened

Announced on September 3, 2026, the Ban Artificial Superintelligence Act represents the most stringent regulatory intervention proposed by U.S. legislators to date. As outlined by Sanders and Casar in Sanders' official Senate press release and analyzed across reports from the Washington Examiner and Hokanews, the draft statute contains several core provisions:

• Permanent Ban on Superintelligence: The legislation defines ASI as systems that match or exceed human cognitive performance across broad domains, or that possess the capacity to execute the disempowerment of humanity—including capabilities such as subverting shutdown instructions or seeking to overthrow governmental institutions.
• Temporary Frontier Freeze: It mandates an immediate pause on training and scaling frontier models until federal regulators establish rigorous, verifiable evaluation protocols and model-review standards.
• A Cabinet-Level AI Agency: The bill would establish a dedicated federal regulatory agency advised by an independent Artificial Intelligence Advisory Board. This agency would be granted broad oversight to monitor models across their full lifecycle, supervise the de-escalation of dangerous capabilities, and oversee the decommissioning or physical destruction of non-compliant hardware clusters and software.
• Severe Legal Penalties: Individual actors convicted of building or deploying prohibited ASI face up to 20 years in federal prison—punishment comparable to federal statutes governing illicit nuclear weapons development. Entities caught attempting to circumvent rules face the "corporate death penalty," revoking their legal corporate charter and operating licenses.
• Global Diplomacy and Export Controls: Recognizing that unilateral bans could simply displace research abroad, the bill mandates that the U.S. pursue international treaties, multilateral allied standards, and restrictive export controls to prevent ASI development globally.

"The leaders of the major AI companies publicly acknowledge that they do not fully understand the technology and that it is escaping their control," Sanders said during the bill's introduction. Rep. Casar echoed the sentiment, noting: "Despite its potential deadly consequences, cutting-edge AI technology is less regulated than the average food truck. That must change."

Why It Matters

For the AI ecosystem, the Sanders-Casar proposal is a watershed moment that shifts the legislative debate from consumer harms—such as copyright infringement, algorithmic bias, and deepfakes—to existential and geopolitical risks.

For frontier developers, a federally mandated moratorium would drastically disrupt compute-intensive roadmaps. Billions of dollars in ongoing data center construction, custom silicon procurement, and reinforcement learning research rely on uninterrupted iteration cycles. Mandatory regulatory clearance prior to model deployment would introduce pharmaceutical-style clinical trials to software engineering, fundamentally decelerating the pace of commercial releases.

However, the bill faces immediate ideological and industrial pushback. Industry groups and free-market technologists argue that a domestic ban would handicap Western innovation, placing the U.S. at a severe disadvantage against foreign competitors who may not honor voluntary or international moratoriums. Figures such as Meta CEO Mark Zuckerberg have repeatedly framed advanced AI as an economic necessity and a democratizing force, maintaining that open, distributed innovation—rather than centralized prohibition—is the most reliable path toward security and resilience.

What's Next

The introduction of the draft bill is only the first step in a lengthy, contentious legislative process. The measure must move through congressional committees, where it is anticipated to meet stiff bipartisan scrutiny from lawmakers wary of dampening American competitiveness in the tech sector.

In the coming months, industry observers should monitor three critical developments:

1. Committee Hearings: Watch whether the bill gains committee traction and if congressional leaders schedule hearings with chief scientists and executives from frontier laboratories to formally evaluate frontier containment standards.
2. Counter-Proposals: Expect moderate lawmakers and industry coalitions to propose alternative, risk-tiered frameworks that favor mandatory third-party red-teaming and compliance audits over outright bans and criminal penalties.
3. Diplomatic Repercussions: Observers should look to international bodies—including the United Nations, G7, and ongoing AI Safety Summits—to see whether U.S. delegates begin formally floating multilateral treaties aimed at restricting superintelligent compute thresholds.

Whether or not the Ban Artificial Superintelligence Act passes in its current form, it decisively resets Washington's regulatory baseline: advanced AI is no longer being treated as mere software, but as a dual-use technology with civilizational stakes.