October 1, 2026. Super intelligence, also written superintelligence, means an intellect that is far better than the best human minds at almost every kind of thinking. Philosopher Nick Bostrom, whose 2014 book Superintelligence made the idea mainstream, defines it as "any intellect that greatly exceeds the cognitive performance of humans in virtually all domains of interest". No system meets that bar today. Since September 29, 2026, the phrase also has a second, official meaning: a US executive order made "Super Intelligence" and "SI" the federal government's words for all artificial intelligence, from chatbots to spam filters. For a business the difference is practical. One meaning describes a future the largest labs say they are racing toward; the other is a new label for the tools you can buy today.
Key numbers
| Item | Number |
|---|---|
| Year Nick Bostrom published Superintelligence | 2014 |
| US order making Super Intelligence the term for AI | September 29, 2026 |
| Legal definition SI inherits (the statutory definition of AI) | 15 U.S.C. 9401(3) |
| Deadline for a proposed federal SI definition (from September 29, 2026) | 60 days |
| OpenAI's 2023 window for expert-level AI in most domains (described as conceivable) | Within ten years |
| Deep Blue beats Garry Kasparov at chess | 1997 |
| AlphaGo beats Lee Sedol at Go | 2016 |
| Signatures on the Statement on Superintelligence (read on October 1, 2026) | 143,562 |
| US adults against superhuman AI until proven safe (polling cited by the statement) | 64 percent |
| US adults who want robust regulation of advanced AI (same polling) | 73 percent |
Definitions, dates and figures read on 1 October 2026 from the executive order, 15 U.S.C. 9401, OpenAI, Meta, Safe Superintelligence Inc. and superintelligence-statement.org.
Super intelligence, defined
The research meaning has three parts. It is general, covering science, strategy, persuasion, planning and nearly everything else, rather than one task. It is far beyond human level, not merely equal to the best experts. And it is usually discussed as artificial superintelligence, or ASI, meaning a machine rather than an enhanced human. Wikipedia's summary calls it a hypothetical agent whose intelligence surpasses that of the most gifted human minds. Writing it as one word or two changes nothing about the meaning, although the two-word form has surged in search since the White House adopted it.
AI vs AGI vs super intelligence
The three research terms describe points on a scale, and the new government usage adds a fourth label that spans all of them.
| Term | What it means | Exists today? |
|---|---|---|
| AI (narrow AI) | Systems built for specific jobs, such as spam filters, recommendation engines, chatbots and AI receptionists | Yes, everywhere |
| AGI (artificial general intelligence) | AI that matches people across most cognitive tasks, not just one | Disputed; labs disagree on where the line sits |
| ASI (superintelligence) | AI that greatly exceeds the best humans in virtually all domains | No |
| SI (US government term) | The federal name for all AI since September 29, 2026, with the same legal meaning as AI | Yes, as a label |
OpenAI drew the same line in its 2023 essay Governance of superintelligence, which describes superintelligence as future AI systems "dramatically more capable than even AGI".
The new meaning: SI in the US government
The September 29 order, Inaugurating the Era of Super Intelligence, tells federal agencies to use "Super Intelligence" and "SI" in place of "artificial intelligence" and "AI" in correspondence, websites, reports and policy documents. It does not claim that superintelligence has arrived. Instead it defines SI as the same technologies covered by the existing legal definition of AI in 15 U.S.C. 9401(3): a machine-based system that makes predictions, recommendations or decisions for objectives set by people. The order argues the new name fits because today's systems increasingly "amplify human ingenuity". Some experts told the BBC the term is unlikely to catch on widely because it normally refers to far more advanced systems. Within 60 days the president's science and technology adviser must propose a formal legal definition of SI, so the official meaning could still narrow. Our report on the order and the White House accord covers who signed what.
Is today's AI super intelligence?
No, not in the research sense. Some systems are superhuman at single tasks: IBM's Deep Blue beat world chess champion Garry Kasparov in 1997, DeepMind's AlphaGo beat Go champion Lee Sedol in 2016, and DeepMind's AlphaFold predicts the 3D shape of proteins, work that once took laboratories months or years per protein. Each is a specialist, which is exactly why none counts as superintelligence. Today's general-purpose chatbots and agents write code, analyse documents and run multi-step tasks, yet they still make basic mistakes and need supervision. The labs' own decisions show how unsettled the frontier is: on September 28, OpenAI scrapped the release of its flagship GPT-6.1 Astra model over safety concerns, the BBC reported.
How far away is superintelligence?
Nobody can give a date, and the boldest forecasts come from the companies building it:
- OpenAI, 2023: it wrote that "within the next ten years, AI systems will exceed expert skill level in most domains" was conceivable.
- Meta, 2025: Mark Zuckerberg's Personal Superintelligence letter says "developing superintelligence is now in sight".
- Safe Superintelligence Inc.: the lab co-founded by former OpenAI chief scientist Ilya Sutskever says on its website that "superintelligence is within reach".
- Yoshua Bengio: the Turing Award winner, commenting on the Statement on Superintelligence, says frontier systems could surpass most individuals across most cognitive tasks within just a few years.
Read these as forecasts by interested parties, not a scientific consensus. Elon Musk, speaking at the White House on September 29, described a future of robots and superintelligence in which everyone could get better medical care than the richest people today, but he gave no timeline.
Why so many people want to slow it down
The fear is not about chatbots. It is about a system smarter than its makers that pursues goals nobody can correct. The one-sentence Statement on Superintelligence calls for a prohibition on developing superintelligence until there is broad scientific consensus that it can be done safely and controllably, and strong public buy-in. Its signatories include AI pioneers Geoffrey Hinton and Yoshua Bengio, and it showed 143,562 signatures when we checked on October 1. The polling it cites found 64 percent of US adults think superhuman AI should not be built until it is proven safe or controllable, or never, and 73 percent want robust regulation of advanced AI. Eliezer Yudkowsky and Nate Soares made the starkest case in their 2025 book If Anyone Builds It, Everyone Dies.
The industry's answer, for now, is self-regulation. Under the White House Accord on Super Intelligence signed on September 29, leading labs pledged four layers of controls: internal monitoring, an internal team that checks it, an independent external auditor and an independent board committee. Critics note the accord is voluntary and has no penalties.
What super intelligence means for your business today
Whatever the label, the AI a business can deploy this year is narrow, measurable and priced per task. An AI receptionist that answers after-hours calls for a plumbing company, an agent that researches leads before a sales call, and a support bot that resolves order questions are all SI under the new federal wording, and none of them is superintelligence. Three habits keep you grounded:
- Judge tools by task results and cost, not by a label. Our breakdown of what an AI agent costs to run shows how to price one.
- Keep a person in charge of decisions that touch money, health or legal exposure, the same principle the accord applies to frontier labs.
- Build for the systems that exist, with logging and escalation, so you can upgrade models as they improve without rebuilding the process around them.
If you want that built for you, our AI automation services start from a workflow audit, not a model choice. And if the SI rebrand has you eyeing a new web address, read our guide to buying a .si domain first.
Frequently Asked Questions
It is an AI that is smarter than the best humans at almost everything, not just at one task. Philosopher Nick Bostrom defines it as any intellect that greatly exceeds the cognitive performance of humans in virtually all domains of interest. No such system exists yet.
AI is any machine-based system that makes predictions, recommendations or decisions for goals people set, which covers today's chatbots and spam filters. Super intelligence, in the research sense, is a hypothetical AI that beats the best humans at nearly every cognitive task. Since a September 29, 2026 executive order, the US government also uses Super Intelligence as its name for all AI, with the same legal meaning as before.
No. AGI usually means AI that matches people across most cognitive tasks. Super intelligence goes further and greatly exceeds the best humans in virtually all of them. OpenAI has described superintelligence as systems dramatically more capable than even AGI.
Not by the research definition. Some systems are superhuman at single tasks, such as chess, Go or predicting protein structures, but none outperforms expert humans at nearly everything.
Nobody knows. OpenAI wrote in 2023 that AI exceeding expert skill in most domains within ten years was conceivable, Meta says superintelligence is now in sight, and Safe Superintelligence Inc. says it is within reach. These are forecasts from companies building the technology, not a scientific consensus.
SI stands for Super Intelligence. Under the executive order signed on September 29, 2026, US federal agencies use SI in place of AI in official correspondence, websites and reports.