Super intelligence (SI) is the name the US federal government adopted on September 29, 2026 for the technology that had been called "artificial intelligence" since the 1950s. If you have asked a chatbot a question, had your phone transcribe a voicemail, or watched a coding assistant finish a function, you have used SI. The word is new; the systems are not. This guide explains what the term means, where it came from, how it differs from the much older idea of artificial superintelligence, and what today's SI systems can and cannot be trusted to do.

The plain definition

In everyday use, SI means software that does things people used to assume required a human mind: answering questions in plain language, drafting and summarizing text, translating, writing code, recognizing speech and images, and making predictions or recommendations from data. Chatbots, image generators, voice assistants and coding agents are all SI tools under the new vocabulary.

There is also a legal definition, and it is deliberately unchanged. The executive order that introduced the term says that, for its purposes, "Super Intelligence" and "SI" mean the technologies covered by the existing federal definition in section 9401(3) of title 15 of the US Code: "a machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations or decisions influencing real or virtual environments." In other words, SI covers exactly the same systems the old term covered. Nothing was added and nothing was taken away.

Where the name came from

The rename happened in three steps over ten days in September 2026.

  • September 22, 2026. In his address to the United Nations General Assembly, President Trump said the US government would use "super intelligence" from then on, arguing: "The use of the word artificial makes intelligence fake. It is not fake. It's actually amazing." The same day, a deputy assistant secretary in the State Department's Bureau of International Organization Affairs told staff to "change all references of 'Artificial Intelligence' to 'Super Intelligence' in our documents and positions."
  • September 29, 2026. The President signed Executive Order 14434, "Inaugurating the Era of Super Intelligence." It directs federal departments and agencies to use "Super Intelligence" and "SI" in official correspondence, public communications, websites, reports and policy documents. Its stated reasoning is that the term "more appropriately captures the promise, potential, and rapidly advancing capabilities of these technologies."
  • October 2, 2026. The order was published in the Federal Register at 91 FR 63129.

For the full text, scope and deadlines, read the Super Intelligence executive order, explained.

How the term is used now

Since the order, the United States has two vocabularies running side by side. Federal agencies are switching: on September 30, 2026, the National Institute of Standards and Technology (NIST) updated its web pages to say super intelligence, and its standards center now appears as the Center for Advancing Innovation and Standards for Super Intelligence (CAISSI). Others have kept the old word. California's governor signed an order on September 30, 2026 telling state agencies to keep saying "Artificial Intelligence," and at the White House event on September 29, Anthropic CEO Dario Amodei and OpenAI president Greg Brockman both used the old term during the press conference.

Laws, contracts and most vendor documentation still use the older wording too, because an executive order cannot rewrite a statute or a signed agreement. AI vs SI: what changed and what didn't walks through the vocabulary, technology and law layers one at a time, and the SI adoption tracker records who has switched and who hasn't, with a link to the original record for each entry.

SI is not artificial superintelligence (ASI)

This is the distinction that matters most. In computer science, *superintelligence* has an older, much narrower meaning: a hypothetical system that far outperforms humans at nearly every intellectual task. The philosopher Nick Bostrom's widely cited 2014 definition is "any intellect that greatly exceeds the cognitive performance of humans in virtually all domains of interest." Nobody has built such a system, and researchers disagree about whether, when, or how one could be built. To keep the two ideas apart, SIBytes writes artificial superintelligence (ASI) when it means the hypothetical one.

Everyday SI is a name for systems that exist today. Calling a chatbot an SI tool does not mean it is superhuman; it means the federal government changed its vocabulary. Not everyone is comfortable with the overlap. Computer scientist Kokil Jaidka said the new name "borrows the prestige of a term researchers reserve for a much bigger leap." The older technical sense is still in active use in Washington, too: on September 23, 2026, Rep. Greg Casar and Sen. Bernie Sanders announced a bill, the Ban Artificial Superintelligence Act, that defines artificial superintelligence in exactly that hypothetical, beyond-human sense.

Between everyday SI and ASI sits a third idea, general SI (AGI): one system that can learn and reason across roughly any task a person can. Our guide to general SI (AGI) covers why nobody agrees on where that line is.

What SI systems are genuinely good at

  • Language. Drafting, rewriting, summarizing, translating, and explaining dense material in plainer words.
  • Breadth. A competent first pass on almost any topic, because these systems were trained on enormous amounts of text.
  • Pattern-heavy work. Writing and debugging code, reformatting data, spotting themes across long documents.
  • Speed and stamina. A tenth draft takes seconds and never tests anyone's patience.

What they still get wrong

  • Facts. SI systems can state false things with total confidence. The failure is common enough to have a name, hallucination, so verify anything that matters.
  • Knowing their limits. They tend to guess rather than say "I'm not sure."
  • Genuinely new problems. Performance can drop sharply on puzzles unlike anything in their training data.
  • The physical world. Robotics is improving, but everyday physical competence remains hard.
  • Memory. Most keep little or nothing between conversations unless a product adds that feature.

Renaming the technology changed none of these strengths or weaknesses. A system that hallucinated on September 28, 2026 hallucinates the same way under its new name.

How to read SI news

  • Ask which sense is meant. A story about "SI tools" or "SI policy" is about today's systems. A story about systems that surpass humans at everything is about ASI, which is still hypothetical. Good reporting makes the difference clear.
  • Check the date. Documents from before September 29, 2026 use the old term for the same technology. A 2025 report on "artificial intelligence" and a 2026 report on SI may be describing identical systems.
  • Don't read the rename as a capability claim. The new term is a vocabulary decision. Look for evidence such as independent evaluations, not labels, when judging what a system can do.
  • Watch the legal wording. Laws, regulations and contracts keep their original language, including the EU AI Act. If a story is about legal obligations, the old term is the operative one.
  • Follow the incentives. Whoever is making a claim may benefit from your believing it. Our guide to reading SI news covers the recurring story shapes and their tells.

Keep going

If you write for work or run a business, the SI terminology guide shows how to adopt the new wording without breaking quotes, legal references or product names. For definitions of every term on this page, see the SI glossary.