superintelligence n. an intellect that greatly exceeds human performance in virtually all domains (Bostrom)
What superintelligence really means
Long before it was a government label, superintelligence was a research idea about machines that outthink people at almost everything.
In research, superintelligence means a hypothetical system that is far more capable than humans across almost every intellectual task. It doesn't exist yet. The US government's "super intelligence" is a different thing: a new official name for today's AI.
The research meaning
The term was popularised by the philosopher Nick Bostrom, who described superintelligence as an intellect that greatly exceeds the cognitive performance of humans in virtually all domains of interest. Researchers often call it ASI, artificial superintelligence, and treat it as a step beyond AGI (human-level general intelligence).
The key word is all. A system that beats people at chess, or writes faster than them, isn't superintelligent in this sense. It has to be better across the board.
Does superintelligence exist?
No. Experts quoted in coverage of the rename were sceptical that today's systems meet the research definition. That's the main criticism of the new label: it borrows the prestige of a term for a much bigger breakthrough.
The government meaning
Since 29 September 2026, US federal agencies use "Super Intelligence" (two words) for the technology the law already calls artificial intelligence. The order doesn't claim today's systems exceed human intelligence. It simply reuses the existing legal definition of AI.
| Research superintelligence | US government "Super Intelligence" | |
|---|---|---|
| Usually written | superintelligence, ASI | Super Intelligence, SI |
| Refers to | A hypothetical future system | Today's AI technology |
| Defined by | Researchers (no legal definition) | 15 U.S.C. ยง 9401(3), for now |
| Exists today? | No | Yes |
Why people worry about superintelligence
Because a system that outthinks people everywhere would be hard to oversee, superintelligence is central to AI safety debates. In the UK, for example, a backbench bill has proposed banning its development; the government has said it won't back a ban but is exploring narrower measures. Our sister site stillcalledai.com covers the UK view.
How to tell which meaning you're reading
- A US agency, the White House, or a company echoing them: it means AI.
- A researcher, safety group or philosopher: it probably means the hypothetical step beyond human-level AI.
- A news headline: read on. Good reporting will say which.