What is SI? Super Intelligence (SI) meaning explained
Last updated · By the SI or Not team · Sources
Short answer
SI stands for Super Intelligence. On 22 September 2026, US President Donald Trump told the UN General Assembly that the US government would call artificial intelligence (AI) "super intelligence" (SI). It is the same technology, such as ChatGPT or Midjourney, under a new name. It is not the academic "superintelligence" (ASI).
SI meaning
In technology, SI means Super Intelligence. It is the term the United States government adopted in September 2026 to replace "artificial intelligence" (AI) in its own documents and communications. When a US official now says "SI", they mean the systems everyone has been calling AI: chat assistants such as ChatGPT, Gemini or Claude, image generators such as Midjourney or Stable Diffusion, and the voice, video and coding models built on the same techniques.
Three things are worth knowing from the start:
- SI is a new label, not a new technology. No model became smarter on 22 September 2026. The software is the same; the word changed.
- SI is not the same as "superintelligence". In research and philosophy, superintelligence (often written ASI, for artificial superintelligence) describes a hypothetical machine that would far surpass human intelligence. That is a different, much bigger idea, explained below.
- SI has other meanings too. If you searched for "SI meaning" because of kilograms and metres, or a sports magazine, jump to other meanings of SI.
On this site we use SI first and AI second, because both terms now describe the same thing. An SI detector is what most people still call an AI detector.
Why AI became SI
The renaming is recent, so here is what the sources listed at the bottom of this page confirm, in order. For the news angle and what the switch changes in practice for readers, companies and newsrooms, read AI is now SI: what the renaming means.
- Days before 22 September 2026Online polls. Donald Trump ran a series of polls on social media about a new name for AI. Press reports differ on the options and results of each round, so we do not repeat figures here.
- 22 September 2026Announcement at the UN. In his speech to the United Nations General Assembly in New York, Trump said: "The use of the word artificial makes intelligence sound fake." He added that "all of United States documents and hopefully the world's, will be changed to use the much more accurate term super, as opposed to artificial." In the same speech he said the United States "totally rejects any attempt to construct a globalist scheme to control artificial intelligence", and that the US was "going to encourage it, not rein it in".
- 22 September 2026The State Department switches to "SI". According to the Associated Press, an email titled "Change in Nomenclature AI to SI" was sent the same day within the Bureau of International Organization Affairs, which runs US missions to the UN and other international bodies. It read: "Following the president's remarks at the GA today, please change all references of 'Artificial Intelligence' to 'Super Intelligence' in our documents and positions." It also said the bureau would use "SI" in all remarks, positions and press elements going forward.
- Following daysFirst use in diplomacy. AP and Euronews report that US representatives used the term "SI" at a UN Security Council meeting on the risks of the technology.
Why the change? The stated reason
The only official reason given is the one in the speech: the word "artificial" makes the technology "sound fake", while "super" is presented as "much more accurate". Commentators have linked the move to other renamings by the same administration, such as the Gulf of Mexico becoming the "Gulf of America" in official US usage in 2025.
What is not confirmed
Scope still unclear. At the time of writing, press reports do not describe a law or a formal executive order behind the new term. AP reported that it was unclear whether other State Department bureaus, or other federal agencies, received the same instruction. Other governments, standards bodies and companies have not adopted "SI", and "AI" remains the dominant term worldwide. We will update this page as official documents appear.
SI vs AI: what changed and what did not
It is easy to read more into a new name than there is. Here is a plain split between what the switch from AI to SI changes and what it leaves exactly as before.
What changed
- Official US vocabulary. US government communications, starting with the State Department's international work, now say "Super Intelligence" or "SI".
- Search and public vocabulary. People now meet "SI" in news, speeches and diplomatic statements, and look up what it means. That is why guides like this one exist.
- The framing. "Artificial" suggested an imitation; "super" suggests something above ordinary. The new word carries a more positive, promotional tone, which the speech made explicit.
What did not change
- The technology. The models, their training, their abilities and their limits are identical under either name.
- The science. Research papers, university courses and technical standards keep using "artificial intelligence" and "machine learning".
- The law elsewhere. Laws written with the term AI, such as the European Union's AI Act, keep their wording and their obligations.
- The risks. Generated images, synthetic voices, hallucinated facts and fake reviews are exactly as common, and exactly as hard to spot, as before.
Rule of thumb: when you read "SI" in a US government context, you can mentally replace it with "AI". When you read "superintelligence" in a book, a research paper or a statement by an AI lab, it usually means something else: machines far smarter than humans.
Super intelligence vs superintelligence (ASI)
The new official term, written as two words, sits right next to an older idea written as one word. They sound the same but they do not mean the same thing.
Superintelligence in the academic sense
The idea of machines that outthink people is older than modern computers, but its best-known formulation comes from the Oxford philosopher Nick Bostrom. In his 2014 book Superintelligence: Paths, Dangers, Strategies (Oxford University Press), he defines a superintelligence as any intellect that greatly exceeds the cognitive performance of humans in virtually all domains of interest. Bostrom distinguishes three possible forms:
- Speed superintelligence: a mind that can do everything a human can, but much faster.
- Collective superintelligence: a system made of many smaller intellects whose combined performance far exceeds any current group.
- Quality superintelligence: a mind at least as fast as a human, but vastly smarter.
In this sense, superintelligence is hypothetical. It is a subject of research, forecasting and debate about safety, not a product you can download. People often call it ASI, artificial superintelligence, to separate it from AGI (roughly human-level general intelligence) and ANI (narrow systems built for specific tasks).
Super Intelligence (SI) as the new name for AI
The US government's "super intelligence" is a relabelling of today's AI. It does not claim that current systems meet Bostrom's definition, and they clearly do not: they make factual errors, need human prompts, fail at tasks that are easy for people and have no goals of their own.
Where the two meanings overlap
The two ideas meet in one place: the current generation of AI, now called SI, is what several labs describe as a step towards more general and eventually superhuman systems. So "SI" names today's tools, while "superintelligence" names a possible destination. Treating them as the same thing either overstates what today's chatbots can do, or understates what the superintelligence debate is about.
AI vs SI vs ASI: comparison table
| AI | SI (Super Intelligence) | ASI (superintelligence) | |
|---|---|---|---|
| Stands for | Artificial intelligence | Super Intelligence | Artificial superintelligence |
| What it is | The established name for today's systems | A new name for the same systems | A hypothetical intellect far beyond humans |
| Exists today? | Yes | Yes (same systems as AI) | No |
| Who uses the term | Researchers, companies, media, lawmakers worldwide | The US government since 22 September 2026 | Philosophers, researchers, AI labs, safety debates |
| Examples | ChatGPT, Gemini, Claude, Midjourney, Stable Diffusion | The same: ChatGPT, Gemini, Claude, Midjourney, Stable Diffusion | None |
| Abilities | Strong on specific tasks, unreliable on others | Identical to AI | Would greatly exceed humans in virtually all domains |
| Key reference | Decades of computer science research | Trump's UN speech, State Department guidance | Nick Bostrom, Superintelligence (2014) |
Examples of SI today
Under the new terminology, every widely used AI product is an SI model. Here are the families you are most likely to meet, grouped by what they produce.
SI chat assistants and language models
- ChatGPT by OpenAI, the chat assistant that brought large language models to the general public.
- Gemini by Google, integrated into Google's search, email and office tools.
- Claude by Anthropic, a chat assistant widely used for writing, analysis and coding.
- Open models such as Meta's Llama or Mistral's models, which developers can run on their own servers.
SI image generators
- Midjourney, known for polished, artistic images.
- Stable Diffusion by Stability AI, an open model that powers many apps and custom tools.
- OpenAI's image models (DALL-E and the image generation built into ChatGPT), Adobe Firefly and Google Imagen.
Other SI models
The same techniques now generate synthetic voices, music and video, and write computer code. Voice cloning and video generation are the source of most deepfakes, which is why they get so much attention.
None of these is a superintelligence in Bostrom's sense. They are narrow systems, even when they seem to handle many topics, and their answers still need checking.
What SI can and cannot do
Whatever you call it, it helps to have a realistic picture of today's systems. This is a general overview, not a benchmark.
What SI models do well
- Draft, summarise, translate and rephrase text in seconds.
- Generate photorealistic or stylised images from a short description.
- Write and explain code, and help find bugs.
- Imitate a voice, a writing style or a visual style from examples.
- Produce content at a scale and speed no person can match, for good or for bad uses.
What SI models cannot do (reliably)
- Guarantee facts. Language models can "hallucinate": state false facts, invent quotes or cite sources that do not exist, with full confidence. Our explainer on why SI hallucinates and how to catch it goes into the causes and real cases.
- Know what they do not know. They rarely signal uncertainty on their own.
- Act with their own goals. They respond to prompts and instructions; they do not decide on their own what to pursue.
- Outperform humans everywhere. They beat people on some tasks and fail at others that people find easy, which is exactly why they are not superintelligent.
- Label their own output. Most generated content carries no visible or permanent mark that it came from SI.
Why detecting SI content matters
That last point is the reason this site exists. A new name does not change the problem: SI models can produce a news photo of an event that never happened, a product review from a customer who does not exist or an essay in a student's voice. Some providers add signed Content Credentials (C2PA) or invisible watermarks, but screenshots and re-saves strip most of them.
An SI detector gives you a second opinion. It reads the statistical traces that generation leaves in pixels or in writing style, plus whatever provenance metadata survives, and turns them into a probability with the clues behind it. SI or Not offers two free tools:
- The SI image detector combines a pixel classifier with EXIF, C2PA Content Credentials and XMP metadata, for photos and renders.
- The SI text detector analyses sentence rhythm, typical SI phrasing and human fingerprints, and highlights the most SI-like sentences.
No SI detector, like no AI detector, is certain. Treat every result as one clue among others, never as proof about a person. If you want to sharpen your own eye first, our guide on how to spot SI images and the SI or Not game are good places to start.
SI glossary
The vocabulary around SI mixes the new official term with older research terms. Here are the ones you will meet most often.
- SI (Super Intelligence)
- The name the US government adopted in September 2026 for what was called artificial intelligence (AI). It covers the same technology: chatbots, image generators, voice and video models.
- AI (artificial intelligence)
- The long-standing name for computer systems that perform tasks associated with human intelligence, such as recognising images, understanding language or generating content. Still the most common term worldwide.
- ASI (artificial superintelligence)
- A hypothetical system that greatly exceeds human cognitive performance in virtually all domains. The academic meaning of "superintelligence", popularised by Nick Bostrom in 2014. No such system exists today.
- AGI (artificial general intelligence)
- A hypothetical system that matches human ability across most intellectual tasks, rather than excelling at one narrow job. Definitions vary between researchers and companies, and there is no agreed test for it.
- ANI (artificial narrow intelligence)
- A system built for a specific task or family of tasks, such as translating text, generating images or recommending videos. Every SI model in use today belongs here, however broad it seems.
- Generative SI (generative AI)
- Models that produce new content, such as text, images, audio, video or code, from a prompt. ChatGPT, Gemini, Claude, Midjourney and Stable Diffusion are generative SI models.
- LLM (large language model)
- A model trained on very large amounts of text to predict the next piece of text. LLMs power chat assistants such as ChatGPT, Gemini and Claude.
- Diffusion model
- The family of image generators, including Stable Diffusion, Midjourney and DALL-E, that build a picture by gradually removing noise from random pixels.
- Deepfake
- Synthetic media, usually a video, image or voice, that imitates a real person. A deepfake is one kind of SI content, not the only one. Read more: What is a deepfake? Types, real cases and how to spot one.
- Prompt
- The instruction or text a user gives an SI model to get an answer, an image or another output.
- Hallucination
- A confident but false statement produced by a language model, such as an invented quote, source or fact.
- SI detection
- Estimating whether a piece of content was made by an SI model rather than by a person or a camera. Also called AI detection.
Other meanings of SI
Most people who type "what is SI" or "SI meaning" are not looking for artificial intelligence at all. If that is you, here are the most common other meanings.
- SI units
- The International System of Units (from the French Système international d'unités), the modern metric system: metre, kilogram, second, ampere, kelvin, mole and candela are its seven base units.
- Sports Illustrated
- The American sports magazine, often shortened to SI.
- Si (silicon)
- The chemical symbol of silicon, the element at the heart of computer chips, written with a capital S and a lowercase i.
- "Sí" and "si"
- "Sí" means "yes" in Spanish, and "si" means "yes" in Italian. Without the accent, "si" means "if" in Spanish and "if" or "yes" in French, depending on context.
- Si in music
- The seventh note of the scale in fixed-do solfège (do, re, mi, fa, sol, la, si), called "ti" or B in English-speaking countries.
- SI in business and medicine
- Depending on the field, SI can also stand for systems integration, a statutory instrument (a type of UK secondary legislation) or sacroiliac, as in the sacroiliac joint.
SI FAQ
What does SI stand for?
In technology, SI stands for Super Intelligence. Since 22 September 2026 it is the term the US government uses instead of artificial intelligence (AI). Outside technology, SI most often refers to the International System of Units (the metric system) or to Sports Illustrated magazine.
Is SI the same as AI?
Yes, as far as the technology is concerned. SI is a new name for the same systems: chat assistants, image generators, voice and video models. What changed is the vocabulary used by the US government, not the software. Most companies, researchers and media still say AI.
Who renamed AI to SI?
US President Donald Trump announced it in his speech to the United Nations General Assembly on 22 September 2026, saying the word "artificial" makes intelligence sound fake. The State Department then told staff in its Bureau of International Organization Affairs to use "SI" in remarks, positions and press material, according to the Associated Press.
Is the renaming of AI to SI a law?
Press reports describe a presidential announcement followed by internal guidance at the State Department. They do not describe a law or a formal executive order, and it was not clear whether other agencies received the same instruction. Other countries are not bound by the change.
What is the difference between SI and ASI?
SI is the new official US name for today's AI. ASI, artificial superintelligence, is a hypothetical future system that would greatly exceed human intelligence in virtually every field, the meaning Nick Bostrom described in 2014. Every SI model in use today is far from ASI.
Is ChatGPT an SI?
Under the new US terminology, ChatGPT is an SI model, because SI is simply the new word for AI. It is not a superintelligence in the academic sense: it makes factual mistakes, has no goals of its own and does not outperform humans across all domains.
Does superintelligence exist today?
No. Superintelligence in Bostrom's sense, an intellect that greatly exceeds human performance in virtually all domains, remains hypothetical. Today's systems can beat people on specific tasks, such as playing chess or drafting text quickly, but they are not superintelligent.
What is the difference between AGI and ASI?
AGI (artificial general intelligence) would match human ability across most intellectual tasks. ASI (artificial superintelligence) would go far beyond it. Both are hypothetical, and researchers disagree about whether and when either could be built.
Why is it called Super Intelligence and not superintelligence?
The US government chose the two-word form "super intelligence", abbreviated SI. The one-word "superintelligence" is an older academic term for machines far smarter than humans, usually abbreviated ASI. The two are easy to confuse, which is why this guide keeps them apart.
Can you tell whether content was made by SI?
Often, but never with certainty. An SI detector looks for statistical traces in pixels or writing style and for metadata such as C2PA Content Credentials. SI or Not offers a free SI detector for images and text that shows a probability and the clues behind it.
Sources
Quotes are taken from the reports below. Where sources disagree on a detail, we leave it out rather than pick one version.
- Trump says he's renaming AI to "super intelligence" · The Washington Post, 22 September 2026
- Trump says he's changing artificial intelligence to "super intelligence" on government documents · Gizmodo, 22 September 2026
- Trump seeks to rename artificial intelligence as "super intelligence" in UN speech · The Hill, 22 September 2026
- Trump tells U.N. he's renaming AI based on a social media poll · The New Republic, 22 September 2026
- Trump: We're going to encourage "super intelligence," not rein it in · CNBC, 22 September 2026
- Trump renames AI "super intelligence" as leaders jostle over dangers · AFP via Yahoo News, September 2026
- US diplomats told to say "super intelligence", not "artificial intelligence", after Trump's call · Associated Press via ABC News, 23 September 2026
- "Artificial" is out: Trump orders officials to call it "super intelligence" instead · Euronews, 24 September 2026
- Livre blanc Super Intelligence (white paper, in French) · Iamin.Digital, 22 September 2026
- Nick Bostrom, Superintelligence: Paths, Dangers, Strategies · Oxford University Press, 2014
Published , last updated .