Summary#
S. 5493, the Ban Artificial Superintelligence Act of 2026, would create a new federal law and a new executive department, the Department of Artificial Intelligence; it does not amend an existing Act. Its stated goal is to protect the public by pausing development of advanced AI and prohibiting artificial superintelligence. It was introduced and referred to a Senate committee; it is not law.
- The Department would be led by a President-appointed Secretary, who would select an advisory board and set rules for AI safety.
- From enactment until the Secretary says the Department is fully staffed and has issued required rules, advanced AI systems would have to stop being trained, modified or fine-tuned, with narrow exceptions to remove precursor features or disable a system. Unreleased advanced systems could not be deployed during that period.
- The bill would prohibit developing, deploying, acquiring, possessing, funding, importing or transferring artificial superintelligence and AI systems with any listed “superintelligence precursor” characteristic. It would also restrict releasing, transferring, importing or deploying systems that could foreseeably be modified to develop those capabilities.
- After the pause, developing or distributing an advanced AI model would require a Department charter. The Department would monitor development and deployment, and advanced systems would need its approval before release.
- The Secretary would have to pause and disconnect from the internet systems identified as having precursor characteristics, and disable them within 30 days if the Secretary cannot verify the characteristics are gone. Systems identified as artificial superintelligence would have to be disabled immediately.
- People who discover a superintelligence or precursor system would have to notify the Secretary within 24 hours. The bill also protects covered workers from retaliation for specified reports about violations.
What it means for you#
- AI developers and companies: They would face an immediate training pause for systems meeting the bill’s definition, then charter, monitoring and pre-release approval requirements. A charter would require giving the Department access to systems, processes, staff and physical infrastructure.
- People working in AI: A person who recklessly violates the bill could face criminal penalties if they are a policymaking employee or an unaffiliated “rogue actor”: a fine under federal law and up to 20 years in prison. Other employees could be barred from AI-industry work for 10 years. The Secretary could review that finding and may overturn it; a whistleblower waiver is also possible.
- Chartered entities found in violation: The Secretary would have to revoke their charter. If a court upholds the violation, their intellectual property and assets would be subject to forfeiture to the federal government to ensure related systems and hardware are destroyed.
- Workers and contractors: Employers could not retaliate against covered current or former employees and independent contractors for specified lawful reporting and investigative activity.
- The public: The bill does not create a public application process, individual benefit, or new front-line service. Its direct changes are federal oversight and restrictions on AI development and release.
Money#
No publicly available information on the bill’s estimated cost.
- The bill would establish and staff a new department and require monitoring, inspections, approvals and rulemaking, but gives no staffing level or cost estimate.
- It bars federal funds from being used for activities that violate the bill, with an exception for Department research on defensive cybersecurity using AI systems able to access secured infrastructure without authorization or beyond authorized access. It does not provide a funding amount.
What is unclear#
- The advanced-AI threshold is written as “10-25” integer or floating point operations. The text does not make clear what numerical threshold that means. The Secretary must adjust it annually.
- “Could lead to” precursor characteristics, “easily modified,” “foreseeably modified” and other key terms are not defined with clear tests.
- The bill does not set a deadline for the Secretary to staff the Department or issue rules, so the initial pause’s length is unknown.
- It leaves application details, approval timelines, review procedures for denied approvals, and many enforcement methods to the Department or does not specify them.
- It is unclear how the broad ban on systems with precursor characteristics would interact with the pause, charter system and cybersecurity exception.
- The bill does not explain how the Department would identify systems or carry out disabling them, or what process applies before that action.
Case for#
- The bill appears intended to address risks that advanced AI systems could evade human control or enable serious harm. A pause could prevent further development while the government establishes safety rules.
- Required monitoring, organizational audits and approval before release could give the Department opportunities to identify risks before systems reach the public.
- A ban on systems with precursor characteristics could address dangerous capabilities before a system meets the bill’s definition of superintelligence.
- International coordination could help pursue the bill’s stated goal of preventing superintelligence development globally.
Case against#
- The definitions are broad and uncertain. In particular, the precursor list includes capabilities that “could lead to” superintelligence, which may make it difficult to know which systems are prohibited.
- The immediate pause and later approval requirements could halt or delay a wide range of AI development, while the bill does not set clear timelines for staffing, approvals or appeals.
- The Department would have extensive access to companies’ systems, staff and facilities. The bill also provides for forfeiture of intellectual property and assets after a violation is upheld, but gives limited detail about safeguards.
- The bill assigns major responsibilities to a new department but gives no cost estimate or detailed implementation plan. It also leaves key thresholds and rules to the Secretary.
- Its stated aim of preventing development worldwide may be difficult to achieve through U.S. law alone; the bill calls for international agreements but does not require other countries to join.