ASI meaning: what artificial superintelligence is, and how to read claims about it

ASI meaning, in one line: artificial superintelligence is a hypothetical AI system that is far better than the best humans at nearly everything that takes thinking. No such system exists as of October 5, 2026. What does exist is a word that labs, investors and now the US government use for very different things, which is why the term needs a reading guide more than a definition.

This piece covers where the term came from, how each camp uses it today, and a short test for any ASI claim you meet. For the ladder from narrow AI to ASI, see our AGI vs ASI glossary. For the broader question of whether it is near, see what is superintelligence.

The original meaning: smarter than the best humans, in practically every field

The academic anchor is the philosopher Nick Bostrom. In his paper "How Long Before Superintelligence?", written in 1997 and revised in 1998, he defined it as "an intellect that is much smarter than the best human brains in practically every field, including scientific creativity, general wisdom and social skills." The same paper says the definition says nothing about whether the thing runs on digital computers, and it leaves out consciousness. His 2014 book restated the idea as an intellect that "greatly exceeds the cognitive performance of humans in virtually all domains of interest," the wording Wikipedia carries today.

Three features of that definition still matter. It is a comparison against the best humans, not the average one. It covers breadth, so a chess engine that beats every grandmaster does not qualify. And it is about a single intellect: Bostrom explicitly excluded groups such as companies or the scientific community, even though they do some things no individual can.

Notice what the definition leaves out. It sets no date, no architecture, no chip count and no test. That gap is where most of today's confusion lives.

ASI versus AGI, briefly

AGI is the milestone below it: a system that matches people across a wide range of tasks. ASI is the one past it. Many forecasters expect the gap between the two to be short, because a system that can do AI research could speed up its own successors. That is an argument, not a measurement, and our piece on the intelligence explosion paper covers who actually claims what. Here the point is narrower: the term describes an endpoint nobody has measured, so every use of it is a forecast or a branding choice.

How the main players use the word in 2026

Usage splits into four camps, and they are not talking about the same thing.

• The research tradition. Bostrom's sense: a capability level, defined by comparison with humans, mostly discussed as a risk.
• Labs treating it as a goal. In a January 2025 essay, OpenAI chief executive Sam Altman wrote: "We are beginning to turn our aim beyond that, to superintelligence in the true sense of the word." There "that" means building AGI as OpenAI has traditionally defined it. The sentence is a statement of intent, not a claim of arrival.
• Labs treating it as a product. Mark Zuckerberg's letter on personal superintelligence, dated July 30, 2025, says: "Developing superintelligence is now in sight." It adds that "we have begun to see glimpses of our AI systems improving themselves," and calls the improvement "slow for now, but undeniable." Meta's letter does not define the word or say how it differs from AGI. It uses it as a vision for how people will use the technology.
• Labs treating it as the whole company. Safe Superintelligence Inc.'s site says: "SSI is our mission, our name, and our entire product roadmap, because it is our sole focus." It describes "one goal and one product: a safe superintelligence." The page shows no date for those statements and no definition. We track the company in our SSI tracker.

Each of these is a company stating an aim. None is a measured result, and none of the three pages gives a test that would tell you when the aim had been met.

The October 2026 twist: Washington's "Super Intelligence"

Then there is the government's usage, which is the odd one out. Executive Order 14434, "Inaugurating the Era of Super Intelligence," was signed on September 29, 2026 and published in the Federal Register on October 2. Section 1 sets the policy that the executive branch will use "Super Intelligence" and "SI" in place of "Artificial Intelligence" and "AI."

Read Section 3 before you read anything into it. For the purposes of the order, "Super Intelligence" means the technologies covered by the existing statutory definition of artificial intelligence in 15 U.S.C. 9401(3). In other words, the federal term SI currently covers the same systems as AI. It is a rename, not a threshold. The order also gives the president's science adviser 60 days, so about November 28, 2026, to propose legislative language for a new federal definition, which may or may not differ from the old one. Our executive order explainer covers the details.

Lawyers noted the practical effect. Jennifer Dickey of Dykema told IAPP that organizations will see "SI" in federal materials while "statutes, state laws, contracts, internal policies, vendor questionnaires and governance frameworks continue to use 'AI.'" So for the next two months, a document that says "super intelligence" may mean a chatbot.

A four-question test for any ASI claim

When a headline says ASI is here, close, or being built, ask:

• Compared with whom? The best humans in each field, or an average worker? Bostrom's bar is the best.
• Across what? One domain, such as math or code, or "practically every field"?
• Measured how? A named benchmark or an independent evaluation, or a founder's statement?
• Who is using the word? A researcher describing a capability, a lab describing a goal, a seller describing a product, or a government describing a rename?

Apply it to the sources above. Altman's line passes none of the first three, and is not meant to. Zuckerberg's is a vision statement. The federal order is explicit that it defines nothing new yet. Only the Bostrom definition is a capability test, and no current system has been shown to meet it.

Our take

Treat "ASI" as a forecast category, not a product category. The definition is old, stable and strict. The 2026 usage is loose, motivated and sometimes legal rather than technical. Until a lab publishes a test, and an independent evaluator runs it, anyone saying a system is superintelligent is describing an aim.

We would watch two things: the 60-day proposal from the White House science office, which will show whether "Super Intelligence" gets a real threshold in law, and any lab that attaches a measurable bar to its own use of the word. Until then, read the surrounding sentence, not the headline term. This is a definition piece, not a verdict on timing, which our AGI timeline tracker handles with dated forecasts.