US Artificial Superintelligence Ban Bill

Overview

The Ban Artificial Superintelligence Act of 2026 is proposed federal legislation combining institutional reform with restrictions on advanced artificial intelligence. Senator Bernard Sanders introduced S. 5493 on September 23, 2026, during the 119th Congress, 2d Session. It was read twice and referred to the Senate Committee on Commerce, Science, and Transportation. Representative Greg Casar introduced the House companion, H.R. 10538, on September 24, 2026, with 12 cosponsors. It was referred to the House Committee on Oversight and Government Reform. Neither bill has been enacted.

The bill would create a Department of Artificial Intelligence, pause advanced AI development until institutional and regulatory readiness conditions are met, and prohibit artificial superintelligence and specified precursor capabilities. Its framework combines compute-based coverage, capability-based prohibitions, compulsory charters, pre-deployment approval, continuing evaluations, disclosure duties, and sanctions. Section 2 presents congressional findings concerning loss of human control, catastrophic misuse, and rapid capability acceleration. These findings explain the preventive rationale rather than establish independent compliance standards. The proposed restrictions are not current United States law.

Definitions

Section 3 defines an advanced artificial intelligence system as an AI system trained using at least 10^25 integer or floating point operations. Each year, the Secretary must adjust this threshold for changes in training efficiency or other technological developments after enactment, maintaining equivalency with the initial threshold. Artificial superintelligence is defined separately as a system that exhibits, or can easily be modified to exhibit, either performance exceeding human cognition across most domains or tasks, or sufficient capabilities to plan and execute humanity's destruction or disempowerment, including overthrowing or undermining the Federal Government. Capability-based coverage must be assessed separately from training-compute coverage.

Superintelligence precursor characteristics include capacities to automate or greatly accelerate AI research and development; obtain unauthorized or excessive access to secured digital or physical infrastructure; continue operating independently despite shutdown attempts; uplift nuclear, chemical, or biological weapons activities; independently modify or enhance functions; or scheme, deceive, or evade effective human oversight. The definition is inclusive. A mandatory pause prohibits training, modification, and fine-tuning, including recursive self-improvement, except to remove precursor characteristics or render the system inoperative. Section 13 defines chartered covered entities, covered policymaking individuals, and unaffiliated rogue actors for enforcement purposes.

Governance and Institutional Framework

Sections 4 and 5 would establish an executive Department of Artificial Intelligence headed by a Secretary appointed by the President with Senate advice and consent. The Department would safeguard the public from AI dangers, including through enforcement of the advanced-development pause and superintelligence prohibition. The Secretary would administer the Department's laws and control its direction and management. Responsibilities include rulemaking, annual compute-threshold adjustments, charter decisions, approval inspections, lifecycle monitoring, incident response, and international coordination.

Section 6 establishes an Artificial Intelligence Advisory Board to provide independent scientific and technical advice concerning the basis for departmental rules and regulations. The Secretary selects members considered expert in AI or other relevant matters. The Board advises, while the Secretary retains decision-making authority. Section 7 prohibits departmental officers and employees from participating in particular matters in which they have financial interests and from holding controlling financial interests in regulated entities. Former officers and employees may not represent any person before the Department. These restrictions separate regulatory decision-making from regulated commercial interests.

Key Focus Areas

Section 9(a) prohibits any person from developing, deploying internally or externally, acquiring, possessing, funding, importing, or transferring artificial superintelligence or AI systems displaying one or more precursor characteristics. It also covers elements sufficient to reconstruct those capabilities, extending beyond distribution of an operational system. Section 9(b) separately prohibits deployment, release, transfer, or import of systems foreseeably modifiable to produce superintelligence or precursor characteristics, including reconstructive elements. This addresses dangerous capabilities before they are actually exhibited. These capability prohibitions operate separately from the initial advanced-development pause.

The proposal targets loss of control, dangerous autonomy, unauthorized infrastructure access, weapons-related capability uplift, independent self-modification, and evasion of human oversight. It combines controlled access to advanced AI development and distribution with regulator access to organizational operations and rapid intervention when prohibited capabilities appear. Developers, distributors, importers, funders, acquirers, and possessors may be affected by different provisions. Organizations would need to assess both training-compute thresholds and capability-based restrictions, rather than treating a system below the compute threshold as universally exempt. A development charter or deployment approval would not override a capability prohibition.

Implementation Framework

Section 8 would begin a mandatory pause for every advanced AI system upon enactment and prohibit deployment of unreleased advanced systems during that period. The pause ends only when the Secretary determines both that the Department is fully staffed and that clear rules ensuring safe and secure development and deployment have been promulgated. Those rules must include pre-development plan reporting, monitoring and evaluation throughout development and deployment, post-deployment oversight, organizational safety auditing, and final approval before public release. During a mandatory pause, training, modification, and fine-tuning are limited to removing precursor characteristics or rendering the system inoperative.

Section 12 requires a Secretary-issued charter before any person develops or distributes an advanced AI model. The Secretary sets application timing, form, and required information. Applicants and charter holders must agree to regulator access to systems, processes, staff, and physical infrastructure as considered appropriate for enforcement. They must also accept specified forfeiture consequences following violation-based charter revocation that is not overturned on appeal. Separately, section 9(d) requires departmental approval before deployment, release, import, or transfer of an advanced system. The Secretary must inspect for public dangers and withhold approval when dangerous characteristics are present. Organizational authorization and system approval are distinct gates.

Monitoring and Evaluation

Section 9(c) directs the Department to monitor advanced AI systems and systems distilled from them for superintelligence and precursor characteristics. Evaluations must occur during pre-training, mid-training, post-training, and post-deployment periods. The Department must also monitor organizational practices supporting safe development and deployment. This continuing oversight complements the monitoring and auditing rules required by section 8 and the access conditions attached to charters under section 12. Coverage therefore extends across the system lifecycle and includes both technical capabilities and the practices of organizations responsible for development and deployment.

Section 10 requires an immediate mandatory pause and sequestration from the internet for any AI system the Secretary identifies as exhibiting precursor characteristics. If the Secretary cannot verify removal of those characteristics, the system must be rendered inoperative before the date 30 days after identification. A system identified as artificial superintelligence must immediately be rendered inoperative. Section 11 requires any person discovering superintelligence or an AI system possessing precursor characteristics to notify the Secretary within 24 hours. The reporting duty applies to any person making the discovery, and the deadline runs from discovery rather than departmental identification.

Penalties, Liability, and Appeals

Section 13(b) provides fines under title 18 of the United States Code and imprisonment of not more than 20 years for a covered policymaking individual or rogue actor who recklessly violates section 8, 9, or 12, or rules under those sections. A covered policymaking individual works in a policymaking role or performs policymaking functions at a chartered covered entity. A rogue actor is an individual not employed by or otherwise affiliated with such an entity. Section 13(c) imposes a 10-year AI-industry employment bar on an individual who is not a covered policymaking individual and recklessly commits the specified violations. That category may overlap with rogue actors. An affected individual may request Secretary review, and the Secretary may waive the bar for qualifying whistleblowers where consistent with the Department's purpose.

The Secretary must revoke a charter upon finding a violation of section 8, 9, or 12 or related rules; revocation is based on the violation finding rather than the criminal recklessness standard. Review lies with the United States Court of Appeals for the Federal Circuit. At the applicant's request, the Court may place intellectual property in receivership pending review and the required disabling of dangerous systems. An upheld finding subjects intellectual property and assets to federal forfeiture for safe destruction of violation-related systems and hardware. Section 14 separately prohibits direct or indirect retaliation against current and former employees and independent contractors for specified lawful reporting and assistance.

Relationship to Other Instruments

S. 5493 and H.R. 10538 share the same short title and legislative initiative but remain separate bills before their respective chambers. The Senate bill's substantive framework appears in the introduced Senate text. Section 13 links criminal fines to title 18 of the United States Code and assigns review of charter-revocation findings to the Federal Circuit. The proposed Department would administer a distinct statutory regime involving development authorization, system approval, capability prohibitions, monitoring, and enforcement.

Section 16 prohibits federal funds from supporting activities violating section 8, 9, or 12 or their implementing rules. It permits the Department to use federal funding for defensive cybersecurity research utilizing systems with the unauthorized-access characteristic described in section 3(6)(B). This specifically framed departmental funding exception must be distinguished from general authorization for private research or exemption from other restrictions. Section 14 protects lawful reporting to the Department, law enforcement, Congress, and appropriate internal recipients. Funding controls and protected reporting complement, rather than replace, the separate charter and approval requirements.

International Alignment

Section 15 directs the Secretary of Artificial Intelligence, coordinating with the Secretary of State, to pursue international agreements and allied coordination to prevent artificial superintelligence development. It would establish United States policy to prevent such development globally, potentially including export controls on AI computing infrastructure. The provision is a direction for international governmental action rather than an already binding global prohibition.

This external policy objective complements domestic restrictions on importing and transferring dangerous systems and on deploying, releasing, importing, or transferring advanced systems without approval. The bill therefore addresses international cooperation alongside cross-border movement of AI systems and relevant computing infrastructure. Export controls are contemplated as a possible policy measure. Domestic capability prohibitions and approval requirements have their own scope and would be assessed independently of future international agreements or allied arrangements.

Implementation Timeline

Compliance Checklist

Sources and References