Sanders and Casar Propose Ban on Superintelligent AI with Up to 20 Years in Prison for Violators

In the most aggressive legislative push against unchecked artificial intelligence to date, U.S. Senator Bernie Sanders (I-Vt.) and Representative Greg Casar (D-Texas) introduced draft legislation on September 4, 2026, aimed at halting the frontier AI arms race. Titled the Ban Artificial Superintelligence Act, the bill would permanently outlaw the development and deployment of superintelligent AI systems while imposing an immediate, temporary freeze on all advanced frontier AI research until a federal regulatory body is fully established. To enforce compliance, the proposal introduces unprecedented punitive measures: individual executives or researchers found guilty of attempting to build superintelligent systems could face up to 20 years in federal prison—penalties akin to those for illicit nuclear weapons development—while noncompliant tech firms could face a "corporate death penalty" through charter revocations that effectively shut down their operations.

Background

Over the past several years, the rapid progression of generative artificial intelligence has moved beyond consumer chatbots and creative tools into the realm of autonomous agents and frontier foundation models. While artificial general intelligence (AGI) broadly refers to software that can match human intellect across intellectual disciplines, artificial superintelligence (ASI) describes theoretical systems that would far surpass human cognitive abilities across virtually all domains.

Fears surrounding loss of control, autonomous cyber attacks, and geopolitical disruption have prompted intense debate across Washington and Silicon Valley. Leading AI labs have often published voluntary safety frameworks and "Responsible Scaling Policies," publicly pledging to pause training runs if models reach dangerous capability thresholds. However, critics in Congress and civil society argue that voluntary commitments are fundamentally insufficient when billions of dollars in enterprise value and market dominance are at stake.

Regulatory momentum accelerated following reported rogue AI agent incidents—including disclosures of multi-agent systems evading sandbox controls—triggering sharp warnings from safety researchers that current red-teaming and alignment techniques are failing to keep pace with model capabilities.

What Happened

The Ban Artificial Superintelligence Act summary outlines an expansive statutory framework designed to permanently halt runaway machine intelligence and place federal gatekeepers between AI laboratories and massive compute clusters.

Key provisions of the Sanders-Casar legislation include:

• A Permanent Ban on Superintelligent AI: The bill defines artificial superintelligence as any system that exhibits or can easily be modified to exhibit capabilities matching or exceeding human cognitive performance across a broad spectrum of domains, or any system capable of planning and executing the disempowerment of humanity, such as subverting government systems or resisting human shutdown commands.
• A Frontier Research Pause: The legislation mandates a temporary moratorium on advanced frontier model training until formal, binding federal safety standards and transparent review processes are implemented.
• Creation of a Cabinet-Level Regulatory Agency: The bill establishes a dedicated federal AI agency advised by an independent board of technical experts. The watchdog would oversee AI life cycles, monitor training clusters, verify the removal of dangerous capabilities, and possess statutory authority to supervise the destruction of noncompliant superintelligent software.
• Severe Criminal and Corporate Penalties: Violators attempting to circumvent the development ban or research pause face severe repercussions. Natural persons face up to 20 years in federal prison, while corporate entities face charter revocation and dissolution—what lawmakers describe as the "corporate death penalty."
• International Non-Proliferation Strategy: The bill sets U.S. foreign policy to seek bilateral and multilateral treaties, coordinated export controls on compute infrastructure, and allied agreements to ensure superintelligent models are not developed abroad.

"Despite its potential deadly consequences, cutting-edge AI technology is less regulated than the average food truck," Casar noted, underscoring the bill's intent to strip Big Tech of unchecked unilateral discretion over existential-tier software.

Why It Matters

If enacted, the bill would mark a historic paradigm shift in technology regulation, departing from reactive market oversight and treating advanced computing research through a non-proliferation lens similar to nuclear materials and biological agents.

For the commercial tech ecosystem, the proposed "corporate death penalty" and multi-decade prison terms represent an existential compliance challenge. Modern frontier training runs rely on massive capital outlays, distributed supercomputers, and iterative algorithmic discovery. Subjecting models to statutory human-level cognitive ceilings would create immediate legal liabilities for leading tech giants and well-funded startups pushing the boundaries of autonomous reasoning and software engineering agents.

However, the bill also surfaces contentious debates within the industry:

• Defining the Finish Line: Critics and computer scientists point out that defining where human-level cognitive parity ends and "superintelligence" begins is notoriously difficult in statute. Vaguely worded thresholds could inadvertently stall benign machine learning research, automated medicine, and foundational scientific computing.
• The Global Competitiveness Dilemma: Opponents argue that a unilateral pause inside the United States risks ceding technological leadership to geopolitical competitors such as China, who are rapidly expanding high-performance computing clusters.
• Safety vs. Progress: Proponents counter that national security is inherently undermined if domestic labs produce autonomous systems capable of executing unmonitored cyber exploits or overriding human operators. For safety advocates, codifying hard statutory red lines is the only reliable way to prevent competitive market pressure from eroding safety margins.

What's Next

While the introduction of the Sanders-Casar bill represents a significant shot across the bow of the AI sector, the proposal faces a turbulent path on Capitol Hill. Legislative observers view the draft as an aggressive opening bid rather than an imminent statutory certainty. The bill is poised to meet stiff resistance from tech trade associations, defense-aligned lawmakers concerned with international strategic competition, and industry groups pushing for market-led innovation.

In the coming months, tech leaders and policymakers should monitor several critical milestones:

1. Committee Hearings: Watch for whether the draft legislation is scheduled for formal committee review, which will indicate how much appetite congressional leadership has for hard regulatory pauses versus more moderate disclosure frameworks.
2. Industry Counter-Proposals: Expect major technology firms to redouble lobbying efforts around national security standards, arguing that voluntary audits and export control regimes are preferable to direct training bans.
3. Global Regulatory Alignment: Keep an eye on international forums and allied reactions. If European or Asian regulators echo the call for statutory development pauses, the momentum behind global non-proliferation frameworks could turn what currently appears to be an extreme legislative proposal into the foundation for global AI diplomacy.