AI Glossary / AI and machine learning / Beginner

Superintelligence ban

Definition

A superintelligence ban is a proposed prohibition on developing, deploying, or otherwise enabling AI systems that exceed a defined superintelligence threshold. The term describes a policy idea rather than one universal rule: proposals differ in scope, duration, enforcement, and conditions for allowing development to resume.

What it means

Superintelligence usually refers in this debate to systems that substantially exceed human cognitive capabilities across a very broad range of tasks. Definitions vary, and a policy must translate the concept into criteria that institutions can actually assess.

A ban can be permanent or conditional. A conditional prohibition might remain in place until specified safety and public-governance requirements are met. A temporary pause with a fixed end date is a different design, although public discussion sometimes uses these terms loosely.

Distinguish advocacy statements, proposed legislation, enacted law, and organizational commitments. Their existence, authority, and practical consequences are different. This entry explains the concept without asserting that a universal ban is in force.

How it works

A workable proposal needs a definition of the covered systems and activities. It must specify who is bound, which jurisdiction applies, what evidence triggers restrictions, and who makes the determination.

It also needs implementation arrangements: monitoring, review, enforcement, exemptions if any, and procedures for disputed classifications. Capability tests, development practices, or resource thresholds may appear in proposals, but none should be assumed to measure superintelligence perfectly.

Conditional proposals require a clear process for deciding whether their release conditions have been met. International coordination can matter because research, computing resources, organizations, and deployed services cross borders.

These are policy design questions. A slogan alone does not resolve technical measurement, institutional accountability, or competing views about acceptable risk and foregone benefits.

Three practical examples

1. Policy reading: A student compares a public statement calling for restrictions with a draft bill, noting which document creates proposed legal machinery and which expresses support.

2. Business planning: A research organization tracks how a particular proposal defines covered activities before assessing possible effects on future projects.

3. Public consultation: Participants examine whether a proposed prohibition has explicit review conditions, independent oversight, and a way to challenge mistaken classifications.

Think of it this way

Think of a proposal to close a bridge until engineers establish that a redesigned structure can be used safely. The closure's purpose, evidence requirements, and reopening authority must all be specified.

The analogy has limits. Superintelligence is not a single visible bridge, its capabilities and risks are disputed, and development can occur across many organizations and countries. Technical assessment and political legitimacy therefore raise different problems from ordinary infrastructure inspection.

Worked example

Consider a fictional conditional policy with two release requirements: an independent scientific process accepts the safety evidence, and a designated public-governance process approves resumption.

There are four simple combinations. If neither requirement is met, the prohibition remains. If only the scientific requirement is met, it remains. If only the governance requirement is met, it remains. Only when both are met is the stated release condition satisfied.

This is an illustration of an “and” condition, not a description of enacted law. Real wording must define who makes each decision, which evidence counts, and whether satisfying the conditions automatically lifts restrictions or merely permits a further review.

A company announcing that its own system is safe would not necessarily satisfy an independently defined scientific process. The identity and authority of the decision-maker are part of the policy.

Illustrative case study

Fictional case: Northstar Research asks an internal planning team to assess a proposed restriction described in a news headline as an “AI ban.”

The team reads the underlying proposal and separates its covered capability level, prohibited activities, review process, and legal status. It discovers that the headline does not explain whether ordinary business applications fall within the proposal's scope.

The planning report therefore records uncertainties and alternative scenarios instead of announcing that all AI projects must stop. The fictional case illustrates careful interpretation of policy documents, not legal advice or a claim about any country's current obligations.

Common misconceptions and mistakes

A call for a ban is not itself an enacted ban. A bill's introductory legislative wording also does not establish that it has become law.

A restriction on superintelligence need not prohibit every AI application. Scope depends on the actual definition and activities covered.

Support for restrictions does not prove that a particular catastrophic outcome is certain; opposition does not prove that risks are negligible.

Do not treat a signature count, a benchmark score, or a company's assurance as a substitute for the specific evidence and authority required by a proposal.

Business use

For businesses, the immediate value of understanding this term is clearer policy monitoring and scenario planning. Track primary documents, their status, geographic scope, and the projects potentially affected.

Separate current obligations from hypothetical future restrictions. Record assumptions about capability thresholds and dependencies on external providers. When a concrete compliance decision arises, obtain jurisdiction-specific interpretation of the actual operative text. Avoid making investment or shutdown decisions from an advocacy headline alone.

People also ask

Does the phrase mean all AI is banned?

No. The intended scope depends on the proposal, and many discussions distinguish superintelligence from narrower tools.

Is a ban the same as a pause?

Not necessarily. Duration, release conditions, and covered activities can differ substantially.

Who decides that a system is superintelligent?

There is no answer built into the phrase. A policy must designate criteria and a decision process, and those choices can be contested.

Check your understanding

A public statement asks governments to prohibit development until two conditions are met. Does signing it create a law or prove those conditions are satisfied?

Answer: No. Signing expresses support for the statement. Legal authority and evidence that conditions are met require separate processes.

Related terms

• AI governance
• Artificial superintelligence
• moratorium
• frontier AI
• capability evaluation
• international coordination
• public consultation
• regulatory proposal

Learn next

Learn to distinguish policy advocacy, draft legislation, and enacted requirements. Compare two proposals by scope, authority, duration, release conditions, and enforcement. Then identify which terms would need operational definitions before an organization could reliably determine whether a project is covered.

More in AI and machine learning

• Structural equation modeling - Structural equation modeling, or SEM, is a statistical framework for representing and estimating relationships among variables through linked equations.
• Structured sparsity regularization - Structured sparsity regularization encourages a model to use a limited set of parameters in patterns that reflect known relationships.
• Sufficient dimension reduction - Sufficient dimension reduction seeks a smaller representation of predictors that preserves all their information about a specified response.
• Supervised Clustering - Supervised clustering groups objects using guidance from known labels.
• Supervised learning - Supervised learning trains a model using examples that pair inputs with known target outputs.
• Support vector machine - A support vector machine, or SVM, is a learning method that can build a decision boundary between classes while balancing separation against errors.

Sources

Future of Life Institute, “Statement on Superintelligence”: https://superintelligence-statement.org/en

Independent review read its two required conditions for lifting the proposed prohibition. United States Government Publishing Office, H.R. 10538, introduced September 24, 2026: https://www.govinfo.gov/content/pkg/BILLS-119hr10538ih/xhtml/BILLS-119hr10538ih.html . Reviewed definitions and sections 8–9 distinguishing a development pause and proposed prohibitions. This is introduced bill text, not evidence of enactment. The worked policy and business case are fictional illustrations; their simplified conditions are not a quotation or legal test.

Previous: Super-resolution Next: Superiority and inferiority ranking method