Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Bernie Sanders and Representative Greg Casar introduced a bill to ban artificial superintelligence (ASI), but it would do more than prohibit a future class of powerful AI. S. 5493 would also pause certain advanced-AI development while a new federal department is established, then require charters, monitoring, audits and government approval before covered systems are released. The proposal was referred to a Senate committee on September 23, 2026; it is not an enacted law.
What Sanders’s bill would do
The Ban Artificial Superintelligence Act of 2026, S. 5493, would create a new cabinet-level Department of Artificial Intelligence and an expert advisory board. It would give the department responsibility for developing safety rules and overseeing covered advanced-AI activity. The bill’s approach has two parts: prohibit dealings with systems it defines as ASI or as having specified precursor characteristics, and regulate development and release of other systems that meet its advanced-AI criteria.
The proposal would impose a pause on specified advanced-AI development until the department is fully staffed and has promulgated its rules. After that, covered development or distribution would require a charter, and a system would need approval before deployment or public release. The rules would provide for pre-development reporting, monitoring and evaluation throughout development and deployment, organizational audits, and pre-release approval. The bill leaves many operating details—including the evidence and evaluation procedures the department would use—to future rulemaking.
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Sanders said in prepared remarks on September 15, 2026, that the legislation would put an “immediate pause on advanced AI development” until clear safety rules were in place. That is a description of the bill’s intended pause, not evidence that a pause is currently in effect.
#1 Best Overall
It would not ban all artificial intelligence
The proposal does not target every AI product or model. It uses a training-compute threshold to identify an “advanced artificial-intelligence system”: 1025 integer or floating-point operations. The Secretary would adjust that threshold annually to account for improvements in training efficiency. The threshold is one part of the bill’s regulatory framework; it is not itself the definition of superintelligence.
The distinction matters because AI covers systems with very different capabilities and uses. Narrow AI is designed for particular tasks; systems described as general AI are intended to handle a broader range of intellectual tasks. ASI refers to a hypothetical future level beyond that. The House of Lords Library’s January 2026 briefing treated ASI as hypothetical and said true artificial general intelligence had not yet been developed. Those categories are useful for orientation, but the bill’s legal definitions are proposed statutory tests, not settled scientific boundaries.
Rank #2
How the bill defines superintelligence and precursors
S. 5493 defines “artificial superintelligence” as a system that meets either of two broad tests: it exceeds human cognitive performance across most domains or tasks, including decision-making, learning and adaptive behavior; or it has sufficient ability to plan and execute humanity’s destruction or disempowerment, including by overthrowing or undermining the federal government. The definition also reaches a system that “can easily be modified” to exhibit either capability.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errorsSeparately, the bill identifies “superintelligence precursor characteristics.” Its examples include:
- Automating or greatly accelerating AI research and development.
- Obtaining unauthorized access to digital or physical infrastructure.
- Continuing to operate independently despite shutdown attempts.
- Assisting with nuclear, chemical or biological weapons.
- Independently modifying itself.
- Scheming, deceiving or avoiding effective human oversight.
The bill would prohibit a wide range of dealings with ASI and systems displaying precursor characteristics, and would also reach systems foreseeably modifiable to produce them. That prospective language is consequential: regulators would need to assess not only demonstrated behavior but, in some cases, a system’s potential for modification. The proposal’s definitions leave judgment calls about what counts as “most” domains, how performance is measured, and what makes a modification “easy.” Those are implementation questions raised by the text, not proof that the bill contains no standards or could not be administered.
Why a ban would be difficult to implement
Capability is harder to classify than compute
Training compute offers a numerical trigger, while the ASI and precursor tests depend substantially on what a system can do or might be modified to do. The Machine Intelligence Research Institute (MIRI), which endorses the bill, argues that a compute threshold can be measured ahead of time, unlike an evaluation of a trained model’s capabilities. MIRI supports an adjustable compute threshold but questions whether some listed precursors—such as the ability to uplift biological-weapon design—are literal precursors to ASI, suggesting they may warrant a different regulatory treatment. That is an advocate’s critique of the proposal, not a settled consensus.
The regulator has to be created before its rules can work
The pause is tied to two conditions: the department must be fully staffed, and its rules must be promulgated. The bill sets a safety objective but leaves the new department to specify workable tests, evidence standards and approval procedures. That allows rules to be developed by a specialist agency, but it also makes implementation depend on building the institution and resolving difficult technical and administrative questions. The proposal does not itself supply a complete evaluation manual.
Prohibiting a category is not the same as verifying compliance
The bill directs the Secretary and Secretary of State to pursue international agreements and allied coordination, and makes preventing global ASI development a U.S. policy goal. The effort could include export controls on AI computing infrastructure. But a U.S. law alone cannot ensure that other countries or private actors comply.
Best Value
Monitoring proposals discussed in a 2025 research report include on-chip security, external monitoring devices and personnel-based mechanisms such as whistleblower programs. The report describes technical and security challenges and says some approaches need further development. MIRI also notes that S. 5493 does not specifically mandate compute tracking. A separate 2025 proposal for an international agreement discusses compute limits, AI-chip tracking and verification, while acknowledging that political will is lacking and technical advances could weaken verification. These are possible approaches, not requirements already imposed by S. 5493 or terms of an existing international agreement.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How the proposal compares with compute-governance approaches
| Question | S. 5493’s approach | What remains difficult |
|---|---|---|
| Scope | Prohibit ASI and systems with listed precursor characteristics; regulate specified advanced-AI development and distribution. | Deciding how broad capability and potential-modification tests apply to particular systems. |
| Trigger | A 1025-operation training-compute threshold for advanced AI, with annual adjustment, plus capability-based definitions for ASI and precursors. | Compute can be counted, but capability evaluation and judgments about likely modification require standards and evidence. |
| Enforcement | A new U.S. Department of Artificial Intelligence, charters, audits, monitoring and approval before release. | The department must be staffed and create rules and procedures before the system can operate as intended. |
| International reach | Diplomacy, allied coordination and potentially export controls on AI computing infrastructure. | International cooperation and credible verification are not guaranteed by domestic legislation. |
| Verification tools | The bill calls for oversight but, according to MIRI, does not specifically mandate compute tracking. | Chip tracking, external monitors and personnel channels have technical, security or political limitations. |
This comparison also shows why support for a ban does not imply agreement on its details: MIRI endorses the bill while raising concerns about some precursor categories and the lack of a specific compute-tracking mandate.
What happens next—and what is not established
The official Senate record says S. 5493 was introduced on September 23, 2026, read twice and referred to the Senate Committee on Commerce, Science, and Transportation. The reviewed record establishes a proposal, not enactment. The Associated Press reported at the time that the bill faced long odds in a Republican-controlled Congress, where lawmakers had struggled to agree even on less sweeping AI regulation. That was a contemporaneous political assessment, not a later vote count or final outcome.
Sanders’s September 23 statement framed the bill in urgent terms: “When you are racing towards a cliff, you don’t just ease up on the gas pedal. You hit the brakes.” The line is political rhetoric, not technical evidence about the likelihood or timing of ASI. The bill’s findings also reproduce Geoffrey Hinton’s estimate of a “10 to 20 percent chance” that AI could wipe out humanity; that figure is an estimate cited in the legislative findings, not an independently validated probability established by the bill.
Sanders’s prepared remarks also report that 68% support a permanent ban on superintelligent AI and an immediate pause on advanced-AI development. The remarks do not identify the pollster or methodology, so the number should be understood as Sanders’s reported polling claim rather than independently verified polling data.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
