Ban Artificial Superintelligence Act of 2026
- Sponsor
- Sen. Sanders, Bernard [I-VT]
- Committees
- Commerce, Science, and Transportation Committee (primary)
- Last action
- Sep 23, 2026
Bottom line
This bill seeks to prevent existential risks from advanced artificial intelligence by establishing a new federal department with broad regulatory and enforcement powers, including a temporary development pause and a permanent ban on superintelligence.
What it actually does
This bill establishes a new executive Department of Artificial Intelligence (DAI) tasked with regulating advanced AI systems. It imposes a mandatory pause on the development and deployment of advanced AI until the DAI is fully operational and has promulgated safety rules. Crucially, it permanently prohibits the development, deployment, acquisition, possession, funding, import, or transfer of artificial superintelligence or AI systems exhibiting superintelligence precursor characteristics, mandating the destruction of any such identified systems.
Proponents argue
Supporters argue that unchecked AI development poses catastrophic risks to humanity, citing warnings from leading experts and industry figures. They contend that a strong regulatory body and a pause on advanced AI development are necessary to ensure safety, prevent the creation of uncontrollable systems, and safeguard civilization from potential destruction or disempowerment by artificial superintelligence.
Opponents contend
Opponents might argue that such a bill would stifle innovation, grant excessive power to a new federal agency, and put the United States at a disadvantage in the global AI race. They might also contend that the definitions of 'superintelligence' and 'precursor characteristics' are too broad or vague, potentially leading to over-regulation of beneficial AI technologies.
The bill is relatively short and clearly structured, making it accessible for an informed reader to understand its core proposals and implications within a reasonable timeframe.
Section 4
Establishment of the Department of Artificial Intelligence
This section creates a new executive department of the United States, to be known as the Department of Artificial Intelligence. Its explicit purpose is to safeguard the public from the dangers of artificial intelligence, specifically by enforcing a pause on the development of advanced artificial intelligence and a prohibition on artificial superintelligence.
Supporters argue
Supporters argue that a dedicated federal department is essential to centralize expertise and authority for addressing the unique and rapidly evolving risks posed by advanced AI, ensuring a coherent and robust regulatory framework.
Critics contend
Critics might contend that creating an entirely new department is an overreach, potentially leading to bureaucratic inefficiencies and an overly centralized approach that could stifle innovation. They might suggest existing agencies could be adapted.
Tradeoffs
The tension between the need for focused regulatory oversight of a novel technology and concerns about creating new, potentially unwieldy government bureaucracy.
Section 8
Mandatory Pause on Advanced AI Development
This section imposes a mandatory pause on all advanced artificial intelligence systems, meaning they cannot be trained, modified, or fine-tuned (except for safety purposes) from the bill's enactment until the new Department of AI is fully staffed and has promulgated comprehensive safety rules. No unreleased advanced AI systems may be deployed during this period.
Section 9(a) and 9(b)
Permanent Prohibition on Artificial Superintelligence and Precursor Systems
This section permanently bans the development, deployment, acquisition, possession, funding, import, or transfer of artificial superintelligence or AI systems exhibiting 'superintelligence precursor characteristics.' It also prohibits deploying systems that could foreseeably be modified to produce such characteristics.
Section 10
Mandatory Sequestration and Destruction of Dangerous AI Systems
This section mandates that the Secretary of AI immediately subject any identified AI system with superintelligence precursor characteristics to a mandatory pause and sequester it from the internet. If the Secretary cannot verify the removal of these characteristics within 30 days, the system must be rendered inoperative. Any identified artificial superintelligence must be immediately rendered inoperative.
Section 12
Charter Requirement for Advanced AI Development and Distribution
This section requires any person developing or distributing an advanced artificial intelligence model to obtain a charter from the Secretary of AI. As a condition of the charter, the person must agree to give the Secretary full access to their AI systems, processes, staff, and physical infrastructure for enforcement. If a charter is revoked due to a violation, and the finding is upheld, the person forfeits their intellectual property and assets for the destruction of related systems.
Section 13
Criminal and Employment Penalties for Violations
This section establishes severe penalties for violations. Covered policymaking individuals or 'rogue actors' who recklessly violate sections 8, 9, or 12 face fines and up to 20 years imprisonment. Other individuals face a 10-year ban from AI industry employment, with a waiver for whistleblowers. Charters can be revoked, leading to forfeiture of intellectual property and assets if violations are upheld.
Defines an 'advanced artificial intelligence system' as one trained on computing power greater than or equal to 10^25 integer or floating point operations, with an annual adjustment by the Secretary to maintain equivalency with technological developments.
Section 3(1)
Why it matters:Standard legislative practice to put definitions upfront, but the technical nature of the threshold and the dynamic adjustment could be overlooked by a casual reader.
Case for: Supporters argue this provides a clear, measurable threshold for regulation while allowing flexibility to adapt to rapid technological changes in AI efficiency, ensuring the regulation remains relevant.
Case against: Critics might argue that setting a specific computational threshold could be arbitrary or easily circumvented by new, more efficient architectures, and that giving the Secretary power to annually adjust it grants too much discretion.
Estimated impact: This definition determines which AI systems fall under the mandatory pause and charter requirements, significantly impacting the scope of the bill's regulation and the operations of AI developers.
Defines 'artificial superintelligence' as an AI exceeding human cognitive performance across most domains or capable of planning/executing humanity's destruction/disempowerment. 'Superintelligence precursor characteristics' include capacities like automating AI R&D, unauthorized infrastructure access, independent operation, WMD uplift, self-modification, and deception.
Section 3(2) and 3(6)
Why it matters:Standard legislative practice for definitions, but the profound implications of these specific definitions for the future of AI development warrant close attention.
Case for: Proponents argue these definitions are necessary to identify and target the specific types of AI that pose existential risks, providing a legal basis for the prohibitions and destruction mandates.
Case against: Opponents contend these definitions are inherently vague, subjective, and difficult to prove or disprove, potentially leading to arbitrary enforcement, stifling beneficial advanced AI research, and creating legal uncertainty.
Estimated impact: These definitions are the lynchpin for the entire regulatory framework, determining what is banned and what triggers mandatory destruction, thereby shaping the entire future of AI development.
As a condition of maintaining a charter, a person agrees that if their charter is revoked due to a violation and upheld on appeal, they forfeit their intellectual property and assets to ensure the safe and complete destruction of all related systems and hardware. This is reiterated as a penalty.
Section 12(d)(2) and Section 13(d)(4)
Why it matters:It's a specific, highly punitive consequence embedded within the broader enforcement mechanisms, rather than a standalone, headline provision.
Case for: Proponents argue that such extreme measures are necessary to ensure compliance and to prevent dangerous AI systems from falling into the wrong hands or being reconstructed, given the existential stakes.
Case against: Critics would argue this constitutes an unconstitutional taking of property, is excessively punitive, and would create an insurmountable barrier to entry and innovation in the AI sector, as the risk of losing all assets is too high.
Estimated impact: Creates an unprecedented level of financial and intellectual property risk for AI companies, potentially driving development offshore or stifling investment in the U.S. AI sector.