Official Definitions of AI and Super Intelligence (SI), Side by Side
A compliance officer re-reads the company’s AI governance policy the week Executive Order 14434 takes effect and asks the only question that matters: does every mention of “artificial intelligence” now have to be rewritten? Our answer is no, and Iternal supports the rename on its merits. President Trump’s Executive Order 14434 names Super Intelligence (SI), formerly called artificial intelligence (AI), as the federal government’s term for the same technology already covered by U.S. law. That is exactly why the rename works: Section 3(a) borrows the existing statutory definition instead of writing a new one, so the name gets sharper while the legal line underneath it does not move. There was never one single official definition of AI to begin with — U.S. statute alone runs two different active definitions at once, and five more official and standards bodies (the OECD, the EU, NIST, ISO/IEC and UNESCO) each wrote their own — so pointing a clearer label at the newer U.S. statutory test is an improvement, not a disruption.
The short answer: seven official and standards-body definitions of “artificial intelligence” are in active use across the U.S. government, the EU, and international standards bodies today, and the newest federal document — Executive Order 14434’s “Super Intelligence” — points back to the existing 2020 statutory definition (15 U.S.C. §9401(3)) rather than replacing it (Source: Federal Register, FR Doc. 2026-20321, Sec. 3(a)). Iternal backs that choice: one accurate name, anchored to the legal line already in force.
A governance document should cite the definition that actually binds it: two U.S. statutory anchors, one executive-order cross-reference, five official and standards-body definitions, and one general-usage dictionary entry for contrast, set side by side below.
There Is No Single Official Definition of AI
The United States alone runs two different statutory definitions at once — a short, capability-based test written in 2020 (Source: 15 U.S.C. §9401(3)) and a longer five-part list Congress wrote two and a half years earlier for a defense-authorization bill (Source: FY2019 NDAA §238(g), Pub. L. 115-232) — and the two executive orders currently in force point to the newer statute while the government’s own agency-use memoranda point to the older one instead (Source: OMB Memorandum M-25-21, p. 18). Beyond the two federal statutes, the center of gravity is a single OECD AI definition that NIST, the EU and (formerly) Colorado each adapted into their own text.
Each definition below sits next to who or what it actually binds. Executive Order 14434 sits second, directly under the statute it borrows from, so the relationship is visible at a glance: it is a cross-reference to the newer and narrower of the two federal statutes, not a new technical test — and Iternal supports President Trump’s rename because the label finally names what the technology does while the legal line it points to does not move.
| Source | Year | Definition (verbatim) | Who/what it binds |
|---|---|---|---|
|
U.S. Statute — 15 U.S.C. §9401(3) Cornell LII, 15 U.S.C. §9401 |
2021 (codified) | “a machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations or decisions influencing real or virtual environments. Artificial intelligence systems use machine and human-based inputs to— (A) perceive real and virtual environments; (B) abstract such perceptions into models through analysis in an automated manner; and (C) use model inference to formulate options for information or action.” | Scopes the National AI Initiative; adopted by cross-reference in Executive Orders 14110 (revoked), 14179, and now 14434. |
|
Executive Order 14434, Sec. 3(a) Federal Register, FR Doc. 2026-20321 |
2026 | “the terms ‘Super Intelligence’ and ‘SI’ mean the technologies and systems encompassed by the term ‘artificial intelligence’ as defined in section 9401(3) of title 15, United States Code.” | A cross-reference, not a new technical test — directs federal executive agencies’ own terminology only; does not reach private companies, state governments, or courts. |
|
U.S. Statute — FY2019 NDAA §238(g) govinfo.gov, Pub. L. 115-232 §238(g) |
2018 (enacted) | “the term ‘artificial intelligence’ includes the following: (1) Any artificial system that performs tasks under varying and unpredictable circumstances without significant human oversight, or that can learn from experience and improve performance when exposed to data sets. (2) An artificial system developed in computer software, physical hardware, or other context that solves tasks requiring human-like perception, cognition, planning, learning, communication, or physical action. (3) An artificial system designed to think or act like a human, including cognitive architectures and neural networks. (4) A set of techniques, including machine learning, that is designed to approximate a cognitive task. (5) An artificial system designed to act rationally, including an intelligent software agent or embodied robot that achieves goals using perception, planning, reasoning, learning, communicating, decision making, and acting.” | The older of the two active U.S. statutory definitions; OMB’s agency-use memoranda M-24-10 and M-25-21 use this one, not 15 U.S.C. §9401(3), for day-to-day federal AI governance. |
|
OECD AI Principles (revised) OECD/LEGAL/0449 |
2023 | “An AI system is a machine-based system that, for explicit or implicit objectives, infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments.” | Non-binding Council Recommendation adhered to by OECD members and non-member adherents; the basis of the 2019 G20 AI Principles, and the base text the EU AI Act and Colorado’s (since-repealed) AI Act both drew from. |
|
EU AI Act, Article 3(1) Regulation (EU) 2024/1689, Art. 3(1) |
2024 (applicable 2025) | “a machine-based system that is designed to operate with varying levels of autonomy and that may exhibit adaptiveness after deployment, and that, for explicit or implicit objectives, infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments.” | Binding law for AI systems placed on the market or put into service in the EU, enforced through a risk-tiered compliance structure. |
|
NIST AI Risk Management Framework 1.0 NIST AI 100-1 |
2023 | “an engineered or machine-based system that can, for a given set of objectives, generate outputs such as predictions, recommendations, or decisions influencing real or virtual environments.” | Voluntary framework, not law — widely referenced by U.S. enterprises and federal contractors building AI governance programs. |
|
ISO/IEC 22989:2022 European Commission, C(2025) 5053 final, quoting ISO/IEC 22989 |
2022 | “reasoning by which conclusions are derived from known premises” (ISO/IEC 22989’s definition of “inference,” as quoted by the European Commission). | ISO/IEC 22989 supplies the AI-system vocabulary underlying ISO/IEC 42001 certification; this definition of “inference” is quoted directly in the European Commission’s 2025 AI Act guidelines. |
|
UNESCO Recommendation on the Ethics of AI UNESCO, 2021 Recommendation |
2021 | “systems which have the capacity to process data and information in a way that resembles intelligent behaviour, and typically includes aspects of reasoning, learning, perception, prediction, planning or control.” | Non-binding ethics recommendation adopted by acclamation by all 193 UNESCO member states; deliberately declines to fix a single technical definition. |
|
Merriam-Webster (general usage) Merriam-Webster |
current | “the capability of computer systems or algorithms to imitate intelligent human behavior.” | General-reference definition — not a legal, regulatory, or technical-standard term of art. |
Why Executive Order 14434 Sharpens the Name Without Moving the Legal Line
President Trump signed Executive Order 14434, “Inaugurating the Era of Super Intelligence,” on September 29, 2026; it was published in the Federal Register on October 2, 2026 (Source: 91 FR 63129, FR Doc. 2026-20321). Section 3(a) does the defining work in a single cross-reference: “the terms ‘Super Intelligence’ and ‘SI’ mean the technologies and systems encompassed by the term ‘artificial intelligence’ as defined in section 9401(3) of title 15, United States Code.” We support that choice. Building the new name on the same statutory test already in force is the clean way to retire “artificial intelligence”, a term that described how the technology was built and was coined in a research proposal dated August 31, 1955, for a two-month summer study at Dartmouth College (Source: A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence, Aug. 31, 1955, reprinted in AI Magazine Vol. 27 No. 4), decades before any of today’s measured results existed — without touching a single line of the statutory test Executive Orders 14179 and 14434 already cite. Super Intelligence, as the order writes it, is not a new capability tier; it is the existing legal definition of AI, named for what it delivers rather than how it originated.
The one date worth a calendar reminder: Section 3(b) directs the Assistant to the President for Science and Technology (APST) to submit, within 60 days of the order — by roughly November 28, 2026 — “proposed legislative language to establish a Federal definition of ‘Super Intelligence’ and ‘SI’.” That proposal must include “an assessment of whether, and to what extent,” the new definition “should modify, expand upon, or otherwise supersede the existing statutory definition of ‘artificial intelligence’” — an assessment, not a decision (Source: Federal Register, FR Doc. 2026-20321, Sec. 3(b)). Until Congress acts on that proposal, 15 U.S.C. §9401(3) remains the operative legal test, and Executive Order 14434’s own “Super Intelligence” definition stays anchored to it.
Two different things change on two different triggers. The statutory definition of AI (15 U.S.C. §9401(3)) changes only if Congress amends the statute. Executive Order 14434’s own “Super Intelligence” definition rests on a different footing: Section 3(a) says it governs “unless and until superseded by subsequent Presidential action consistent with applicable law or by an Act of Congress” — so a future President could redefine “SI” without Congress, even though the AI statute it currently points to would stay the same (Source: Federal Register, FR Doc. 2026-20321, Sec. 3(a)).
The order’s reach matches its purpose: a federal-communications standard, not a private mandate. Section 2(a) directs agencies to use “Super Intelligence” and “SI” “in official correspondence, public communications, websites, reports, policy documents, and other non-statutory documents within the executive branch” — and Section 2(b) adds that nothing requires altering previously issued regulations, contracts, or other historical documents (Source: Federal Register, Sec. 2(a)-(b)). Like the same administration’s earlier “Department of War” secondary title for the Department of Defense (Source: Executive Order 14347, Federal Register pre-publication, Sec. 2), that language reaches federal agencies’ own communications. It does not reach private companies, state governments, or courts, which are free to keep saying “artificial intelligence” — and Iternal’s own choice to say “Super Intelligence” is made on the merits, not under any legal requirement to do so.
Where the Definitions Actually Disagree
The federal government runs two definitions side by side. Executive Orders 14110 (revoked), 14179, and now 14434 all point to the short 15 U.S.C. §9401(3) test. The Office of Management and Budget’s own agency-use memoranda — M-24-10 and M-25-21, issued by opposite administrations about a year apart — both reach past that statute instead and adopt the longer, five-part FY2019 NDAA §238(g) definition (Source: OMB M-25-21, p. 18). Super Intelligence inherits that same split: Executive Order 14434 ties SI to 9401(3), leaving OMB’s separate, broader definition for day-to-day agency AI governance untouched (Source: EO 14434, Sec. 3(a)).
The EU AI Act definition of AI triggers risk tiers; the U.S. statutes don’t regulate at all. The EU AI Act’s Article 3(1) definition exists to trigger a compliance structure — prohibited practices, high-risk obligations, transparency duties — layered on top of the definition itself. Neither 15 U.S.C. §9401(3) nor the FY2019 NDAA §238(g) definition does that; both only scope which federal investments, reports, or programs count as “AI,” with no risk tiers or obligations attached to the definition itself.
The NIST AI definition frames trustworthiness; the statutes frame scope. NIST adapted the OECD’s original 2019 sentence into the AI Risk Management Framework, while the EU AI Act tracks the OECD’s revised 2023 text instead — two different snapshots of the same evolving OECD language. NIST then builds an entire voluntary trustworthiness structure on top of its definition: systems should be valid and reliable, safe, secure and resilient, accountable and transparent, explainable and interpretable, privacy-enhanced, and fair with harmful bias managed — a structure that has no equivalent in either U.S. statute (Source: NIST AI 100-1, Section 3).
ISO supplies the vocabulary other standards build on. ISO/IEC 22989 is the connective tissue under several of these definitions — the European Commission’s own 2025 AI Act guidelines quote its definition of “inference” directly to interpret the Act (Source: European Commission, C(2025) 5053 final).
Two bodies declined to fix a single definition of AI. UNESCO wrote in 2021 that a fixed definition “would need to change over time,” and the UK’s 2023 white paper found “no general definition of AI that enjoys widespread consensus,” choosing instead to regulate two characteristics — adaptivity and autonomy (Source: UK CP 815, §3.2.1). The U.S. and EU made the opposite bet: legislate exact wording now, and let an executive order or a 60-day legislative proposal update it later.
Set against that much existing disagreement, Executive Order 14434 adds none of its own. It does not change which definition OMB uses for agency AI governance, does not touch the EU or NIST text, and does not relitigate what UNESCO or the UK declined to fix. It borrows the newest U.S. statutory test and gives it a name that describes the capability instead of the 1955 Dartmouth proposal that coined “artificial intelligence.” That is why Iternal supports it: a clearer label, pointed at a statutory definition already in force.
SI Is a Statutory Label, Not Bostrom’s Theoretical Threshold
The most common objection to the rename is a definitional one: today’s systems are not “superintelligent” in the sense philosopher Nick Bostrom defined the word in 2014 — “any intellect that greatly exceeds the cognitive performance of humans in virtually all domains of interest” (Source: Nick Bostrom, Superintelligence: Paths, Dangers, Strategies, 2014, Ch. 2, via Wikipedia, "Superintelligence"). That is a fair point about a different definition, and it is not the one Executive Order 14434 uses. Section 3(a) defines “Super Intelligence” and “SI” as “the technologies and systems encompassed by the term ‘artificial intelligence’ as defined in section 9401(3) of title 15, United States Code” — the existing, capability-based statutory definition already in federal law, not Bostrom’s theoretical threshold. Read side by side, the objection answers itself: SI, as Executive Order 14434 defines it, is not ASI as Bostrom theorized it, and Section 3(a) makes no such claim.
Bostrom made a version of this same point directly when asked about the rebrand. “AI” has traditionally “been this wide umbrella term that refers to every form of intelligent information processing in machines,” he told CNN, adding that stretching “superintelligence” to cover all of that makes it “a little bit of a stretch, maybe, to call my spam filter a superintelligence” — while noting “it’s not up to me to decide how English language should be used” (Source: CNN, Zachary B. Wolf, "Trump really wants to make ‘super intelligence’ happen. Is it going to happen?", Oct. 11, 2026). He is right that the word is being asked to cover a wide range of systems. The statute settles exactly how wide: whatever 15 U.S.C. §9401(3) already covers, and no wider. If a spam filter meets that test, Super Intelligence is simply its federal name; the label makes no claim that the filter clears Bostrom’s bar.
None of this reaches outside the federal government, and Iternal is not claiming it does. Section 2(b) of the order states plainly that “nothing in this section requires the alteration of previously issued regulations, Presidential actions, contracts, grants, or other historical documents,” and Section 4(c) creates no enforceable private right or benefit (Source: Federal Register, FR Doc. 2026-20321, Sec. 2(b) & 4(c)). Iternal uses “Super Intelligence” because the name is accurate, not because any statute compels it — and the same choice is available to any organization whose compliance documentation does not change today.
Other governments and legislators are free to choose differently, and the statutory line does not move either way. California’s Executive Order N-10-26, signed the day after President Trump signed EO 14434, directs that “all agencies and departments subject to my authority shall refer to Artificial Intelligence and AI as ‘Artificial Intelligence’ and ‘AI,’ notwithstanding any rebranded or different terminology used by the federal government, unless inconsistent with the law” (Source: Office of Governor Gavin Newsom, signed Executive Order N-10-26, Sept. 30, 2026). In Congress, the sponsors’ summary of the Ban Artificial Superintelligence Act of 2026 (H.R. 10538) defines “Artificial Superintelligence” as a system that “exhibits or can easily be modified to exhibit capabilities that match or exceed human cognitive performance and capabilities across a broad range of domains or tasks,” or one with “sufficient capabilities to plan and execute the disempowerment of humanity” (Source: Sen. Bernie Sanders’ office, Ban Artificial Superintelligence Act of 2026 release summary) — again, a different, higher bar than Section 3(a)’s. A federal order, a state order and a House bill do three different jobs, and none of them amends 15 U.S.C. §9401(3). Section 3(a) is the only one of the three that points a name at that statute, which is why the rename sharpens the label without moving the legal line.
Most Americans still prefer the older term. YouGov surveyed 21,449 U.S. adults on September 28, 2026: 53% called “artificial intelligence” the better name, 9% chose “super intelligence,” 23% had no preference and 15% were not sure (Source: YouGov, "Which of the following do you think is a better name for computer intelligence technology?", Sept. 28, 2026). That is why Iternal writes “Super Intelligence (SI), formerly called artificial intelligence (AI)” on first mention while it leads the shift: the bridge carries readers to the new name, and it is a reason to bridge, not a reason to wait. For the capability tiers in detail, see how SI, AGI, and ASI actually differ.
What This Means for Your Compliance Documentation
Cite the standard your program already follows. A compliance document aligned to NIST’s AI Risk Management Framework, the EU AI Act, or ISO/IEC 42001 does not need a new definition because of Executive Order 14434 — none of those three documents changed. Name the specific standard and version for the legal citation, and use “Super Intelligence (SI), formerly called artificial intelligence (AI)” in the surrounding prose: the citation stays exact and the terminology stays current.
Nothing forces a rewrite; Iternal made the switch on the merits. Executive Order 14434 binds federal executive agencies’ own correspondence and websites — Section 2(b) leaves your existing contracts, regulations, and historical documents untouched, so no contractor, business partner, or customer-facing policy has to adopt the new wording on any deadline. Iternal updated its own public communications to “Super Intelligence”, because the name is the more accurate one for what these systems now deliver — not because a statute required it. Update your forward-looking documentation on the same logic, on your own timeline, and leave anything that cites a dated standard exactly as it is.
Federal contractors should watch agency wording. An organization selling directly into federal agencies should track how individual agencies apply Section 2’s “Super Intelligence” label in RFPs, contracts, and solicitations. The underlying technical and compliance requirements are unchanged — which is the point: a sharper name for the same accepted standard, not a new bar to clear. See the federal-contractor-specific guidance below for the wording patterns to expect, and Iternal’s broader guide for AI in government contracting for the compliance fundamentals that don’t change.
Need AI that already clears federal and regulated-industry compliance bars?
AirgapAI is approved for HIPAA, CMMC, and FedRAMP use cases — so the terminology question stays separate from the compliance question. See how Iternal maps AI governance to federal requirements.
See FedRAMP AI ComplianceSelling into federal agencies right now?
Super Intelligence (SI) for federal contractors and government buyers covers the exact RFP and solicitation wording to expect while Executive Order 14434’s terminology rolls out.
Read the Federal Contractor GuideFAQ: Official AI and Super Intelligence (SI) Definitions
The U.S. government runs two active statutory definitions at once. 15 U.S.C. §9401(3), written as part of the National AI Initiative Act of 2020 and codified January 1, 2021, defines AI as a machine-based system making predictions, recommendations, or decisions from human-defined objectives, then lists three things such systems do with their inputs — the definition Executive Orders 14179 and 14434 both use. The FY2019 NDAA §238(g), enacted in 2018, defines it instead as a five-part list — the definition OMB’s own M-24-10 and M-25-21 agency-use memoranda use.
No, and that is exactly why the rename works. President Trump’s Executive Order 14434 defines “Super Intelligence” and “SI” in Section 3(a) as the same technologies already covered by 15 U.S.C. §9401(3) — a sharper name pointed at the existing statutory test, not a new one. Section 3(b) separately gives the Assistant to the President for Science and Technology 60 days (to roughly November 28, 2026) to propose legislative language that could eventually change the underlying statute, but that proposal still needs Congress to act before any new statutory definition takes effect. The executive order’s own “Super Intelligence” definition is different: it governs unless and until a future President supersedes it or Congress acts, so a future President could change it without Congress. Iternal supports the order on those terms: the label gets clearer, the legal line does not move.
Article 3(1) of Regulation (EU) 2024/1689 defines an “AI system” as a machine-based system designed to operate with varying levels of autonomy, that may adapt after deployment, and that infers from its inputs how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments. This definition became directly applicable on February 2, 2025, and triggers the Act’s risk-tiered compliance obligations.
NIST AI 100-1 (January 2023) defines an AI system as “an engineered or machine-based system that can, for a given set of objectives, generate outputs such as predictions, recommendations, or decisions influencing real or virtual environments,” adapted from the OECD’s 2019 language and ISO/IEC 22989. It is voluntary, not law, but is widely referenced by U.S. enterprises and federal contractors building AI governance programs.
No. At least seven official and standards-body definitions of artificial intelligence are in active use today across U.S. statute, the EU, and international standards bodies. UNESCO is one of them: its 2021 Recommendation on the Ethics of AI deliberately declined to fix a single definition, arguing it would age out as the technology changes. The UK government made a similar choice in its own 2023 white paper, regulating two characteristics — adaptivity and autonomy — instead of adopting a fixed definition. That fragmentation is exactly why Iternal welcomes Executive Order 14434 pointing “Super Intelligence” at one clear, already-in-force anchor instead of adding an eighth definition to the pile.
Because “Super Intelligence” and “superintelligence” are two different terms with two different definitions. Philosopher Nick Bostrom defines superintelligence as “any intellect that greatly exceeds the cognitive performance of humans in virtually all domains of interest” (2014) — a high, theoretical bar. Executive Order 14434 Section 3(a) defines “Super Intelligence” and “SI” by a different, lower one: the existing statutory test at 15 U.S.C. §9401(3), which covers the capability already in use today. The order never claims today’s systems meet Bostrom’s threshold; it names them by the statute that already governs them.
Executive Order 14434 gives the Assistant to the President for Science and Technology 60 days from the order’s September 29, 2026 signing — roughly November 28, 2026 — to submit proposed legislative language. That proposal would still need to pass Congress before it changed the underlying statutory definition of AI, so no finalized statutory definition exists yet. The executive order’s own cross-reference definition of “Super Intelligence” stays in effect in the meantime, unless a future President supersedes it first.
The Statute Didn’t Move. The Name Finally Caught Up.
Seven official definitions of artificial intelligence, one executive-order cross-reference, and one new label pointing back at the newer of the two U.S. statutory definitions. We support President Trump’s Executive Order 14434 and its choice of Super Intelligence (SI): it directs federal agencies’ own communications (Section 2(a)), leaves existing regulations and contracts unaltered (Section 2(b)), and creates no enforceable private right or benefit (Section 4(c)). It also replaces a 1955 research-proposal label about how the technology was built with one that names what it delivers today. Cite the standard your governance program already follows, update your own terminology where accuracy calls for it, and read the full case in Why Iternal Supports Super Intelligence.