Super Intelligence (SI) is, as of 29 September 2026, the term the US federal executive branch uses for what everyone else calls artificial intelligence. The executive order that introduced it defines SI as the technologies and systems already covered by "artificial intelligence" in 15 U.S.C. §9401(3).
That definition is worth reading twice, because it settles the question people keep asking: SI is not a new or more advanced category of technology. It is the same set of systems under a different name.
Two meanings of one phrase
The confusion is understandable, because "superintelligence" already meant something specific.
In the academic and safety literature, superintelligence refers to a hypothetical system that substantially exceeds human cognitive performance across essentially all domains. It is a forward-looking concept, and no system available today meets it.
In US federal usage since 29 September 2026, Super Intelligence means the technologies that were called AI the day before — machine learning models, language models, computer vision, the classifier in your document system.
So a US agency writing "SI" is not making a claim about capability. If a vendor writes it and implies the first meaning, that is a marketing decision, not a regulatory one.
What counts as SI in practice
Since the definition points at the existing statutory definition of artificial intelligence, the practical scope is what it always was:
- Machine learning models trained on data, including the static, deterministic kind that regulated environments prefer.
- Large language models and generative systems that draft or summarise text.
- Classification, prediction and pattern-detection systems — the GAMP classifier, the deviation triage tool, the drift monitor.
- Computer vision, including inspection and imaging applications.
A rules engine or a deterministic calculation is not SI, for the same reason it was never AI. The distinction that matters in a regulated setting is not the label but whether the system's behaviour is reproducible and explainable — which is why GxP AI validation starts with intended use rather than with technology.
Does the word change anything for you?
No, with one caveat. The order is a terminology instruction to federal agencies covering non-statutory materials. It does not amend regulation, and it explicitly does not require previously issued regulations to be altered. Your obligations under Part 11, Annex 11, the draft Annex 22 and GAMP 5 are exactly what they were.
The caveat: a federal definition of SI has been commissioned and not yet published. If that definition eventually diverges from the current one — narrowing or widening what counts — it could matter. Until it exists, treat SI and AI as synonyms, because the order says they are.
More in what the rename changes for GxP and GxP AI. Definitions of the surrounding terms are in the GxP glossary.
Where to go next
Explore GxP Copilot for AI-native validation, TraceDraft for source-traceable clinical documentation, or book a demo to see either on your own data.