Washington Writes ‘Super Intelligence’ Into the Rulebook

Key points

• Executive Order 14434 of 29 September 2026 defines “Super Intelligence” as the technologies already covered by the statutory definition of artificial intelligence in 15 U.S.C. 9401(3), and requires no change to existing regulations, contracts or grants.
• Between 26 and 30 October the term appeared in a housing executive order, a final health-IT rule effective 28 December 2026 that removes the “model card” source-attribute requirements from 45 CFR 170.315(b)(11), a Treasury proposed rule (comments due 7 December) that would prohibit United States investment in the development of any “SI system” in a country of concern, a Department of Energy critical-material listing, a Coast Guard research notice and an Education Department final rule.
• The President’s science adviser must submit draft statutory language defining “Super Intelligence”, with any conforming amendments to existing statutes, within 60 days of the order; the order does not say whether that proposal will be published.

Between 26 and 30 October 2026 the United States government published at least six official instruments that say “Super Intelligence” or “SI” where comparable documents said “artificial intelligence” a month earlier. One, Executive Order 14438 on housing costs, makes it administration policy to “embrace innovative risk and value assessment technologies that incorporate Super Intelligence” in mortgage underwriting and valuation. One, the Department of Health and Human Services’ final rule on health-IT certification, uses the new term for a transparency obligation it is deleting. One, the Treasury’s outbound-investment proposal, uses it for an investment it would ban outright. For a ministry, a regulator or a vendor outside the United States the question is not what the word means, because the order that coined it says it means what “artificial intelligence” already meant. The question is what each instrument does under the new name, and what happens to every contract clause and register written against the old one when a statutory definition arrives.

What the order did: a new label with the old definition attached

Executive Order 14434, “Inaugurating the Era of Super Intelligence”, was signed on 29 September 2026 and published in the Federal Register on 2 October (91 FR 63129). Its purpose section argues that today’s systems “far exceed what was envisioned when the term ‘Artificial Intelligence’ first came into use”, and sets a policy that the executive branch “shall use the terms ‘Super Intelligence’ and ‘SI’ in place of ‘Artificial Intelligence’ and ‘AI’ and will not acknowledge the usage of ‘Artificial Intelligence’ and ‘AI’ in any applicable setting”. Section 2 applies that to correspondence, websites, reports, policy documents and “other non-statutory documents”, “to the maximum extent permitted by law”.

Three limits matter more than the policy. Section 2(b) states that nothing in that section “requires the alteration of previously issued regulations, Presidential actions, contracts, grants, or other historical documents”. Section 3(a) defines the new terms as “the technologies and systems encompassed by the term ‘artificial intelligence’ as defined in section 9401(3) of title 15, United States Code”, the definition enacted in the National Artificial Intelligence Initiative Act of 2020. That statutory definition covers “a machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations or decisions influencing real or virtual environments”. And section 3(b) gives the Assistant to the President for Science and Technology 60 days to submit “proposed legislative language to establish a Federal definition” of the new terms that reflects the capabilities described in the purpose section, with an assessment of whether it should “modify, expand upon, or otherwise supersede” the statutory one and “any proposed conforming amendments” to existing statutory references to artificial intelligence. Sixty days from 29 September is 28 November.

The order has no penalty or enforcement provision and imposes no obligation on private parties, as a 13 October analysis by the law firm Arnold & Porter notes; the same analysis observes that “only Congress can change the statutory definition” and that the order “does not say whether the proposal will be made public”. The renaming began before the order was signed. On 23 September the Associated Press reported an email from Michael Drager, deputy assistant secretary of state in the Bureau of International Organization Affairs, headed “Change in Nomenclature AI to SI”, sent on 22 September shortly after the President told the UN General Assembly that “the use of the word artificial makes intelligence fake”. A department-wide cable followed. Nextgov/FCW reported on 20 October, from an unclassified cable obtained with its sister title Government Executive, that “all official titles and the names of Department governance bodies, systems, platforms, inventories, and programs must use SI in place of AI”, with exceptions for language required by law, legal citations, previously published documents and foreign organisations’ names; contracts and grants were excluded “pending further guidance”. The cable is not public and that account rests on Nextgov/FCW alone. AI & GPP noted it in last week’s Briefing; since then the term has moved from communications guidance into instruments with legal effect.

Where the word now sits, and what each instrument does under it

The table is built from the primary texts; its third column asks not what each instrument says about “SI” but what it does.

Two of these change what someone must do. The health-IT rule is a subtraction. The final rule published on 28 October (RIN 0955-AA09, document 2026-21943) says that one of its larger revisions narrows the “decision support interventions” certification criterion “to fully remove the super intelligence ‘model card’ requirements” in 45 CFR 170.315(b)(11)(iv), for source-attribute support, and (b)(11)(v), for source-attribute access and modification. The amendatory text also removes (b)(11)(vi), the intervention risk-management requirement for predictive decision-support tools, and the matching maintenance duty in 170.402(b)(4). The rule’s own description of what is going is plain: the source attributes “essentially function as nutrition labels”, telling a clinician how a predictive tool was trained, what it is meant for, where it falls short and how its results vary between groups of patients. ONC, the Office of the National Coordinator for Health Information Technology, records that “a large majority of commenters were not supportive”, that many thought removal premature because the requirements had been in force for less than two years, and that commenters feared a “regulatory gap” in SI governance. Its answer rests on developers’ evidence that “most clinical users lack the time or technical background to engage with source attribute information”, on finding no evidence the requirements had improved care, and on being “not persuaded of the value of the transparency requirements”. Whether elements of “source attributes transparency” should be restored will be “taken under consideration for future rulemaking”. The rule is effective on 28 December 2026 (the (b)(11) removals are not among the three amendatory instructions deferred to 1 January 2027). ONC puts the present value of the whole rule’s savings at $1.55bn, or $108.1m a year, across more than 450 certified developers, and refers to the “Office of the HHS Chief Super Intelligence Officer (previously Chief Artificial Intelligence Officer)”.

The Treasury rule is an addition. The proposed Outbound Investment Security Program Regulations (91 FR 67396, document 2026-21786, docket TREAS-DO-2026-0595) would implement the Comprehensive Outbound Investment National Security Act, enacted in the FY2026 National Defense Authorization Act, through a new 31 CFR Part 851. The preamble is candid that the most notable change to the existing sectors “involves what was defined as AI systems in the Part 850 Regulations and is referred to as ‘super intelligence systems’ (SI systems) in the Proposed Rule”, and a footnote cites section 2 of EO 14434 as the reason. The substance is not cosmetic. Under the Part 850 regime in force since 2 January 2025, an AI investment was prohibited, notifiable or outside scope depending on training compute and end use. The proposal drops that: “no single measure, benchmark, or threshold is currently sufficient to distinguish advanced models”, so the Secretary has determined to “define prohibited technology to include the development of any SI system” by a covered foreign person in China (including Hong Kong and Macau), Cuba, Iran, North Korea, Russia or Venezuela under the Maduro regime, which Treasury notes “no longer exists”. Proposed section 851.232 defines an SI system as a machine-based system that can “generate content or make predictions, recommendations, or decisions influencing real or virtual environments”, and sweeps in any software, hardware or tool operating on one, “including a harness for agentic capability or an embodied SI system”. Comments close on 7 December 2026. The law firm Sidley Austin’s 28 October note calls the draft “extremely broad”, reads it as prohibiting investment in companies developing almost any SI system “regardless of end use or level of computational training”, and expects the final regulations to take effect around late March 2027. Treasury’s own press release of 22 October shows the seam: its list of the five statutory sectors says “artificial intelligence systems”, because the statute does, while its description of the proposal says “super intelligence systems”.

The remaining instruments carry the word without changing any duty. The housing order uses “Super Intelligence” once, in its purpose clause; the operative sections direct the Federal Housing Finance Agency to consider a pilot that cancels mortgage insurance using “automated valuation models and other available data”, and a nationwide programme accepting “automated title review or other title-risk assessment” in place of lender’s title insurance on refinances. The Coast Guard’s notice (docket USCG-2026-0374) says the service intends to sign a cooperative research and development agreement with a firm called LOOKOUT to test “Super Intelligence for automated target detection” in search-and-rescue sensor feeds; such agreements “are not procurement contracts”, carry no federal funding, and rival five-page proposals are invited by 25 November. The Department of Energy’s helium notice, signed on 21 October, lists helium because advanced semiconductor manufacturing “is essential for SI and data center supply chains”. The Education Department’s accreditation rule of 30 October mentions SI only in reply to one commenter, saying that regulating it in educational assessment is beyond the rule’s scope and that nothing in the rule prohibits its use, with a footnote: “Pursuant to Executive Order 14434, 91 FR 63129 (Oct. 2, 2026), this document uses the term ‘Super Intelligence’ or ‘SI’ instead of ‘Artificial Intelligence’ or ‘AI.'”

The same sentence under two names: a test of what the rename changes

The cleanest test is the health-IT rule, because it exists in two versions on either side of the order. ONC’s proposed rule of 29 December 2025 (90 FR 60970), published under the office’s former name ASTP/ONC, promised to narrow the decision-support criterion “to fully remove the artificial intelligence (AI) ‘model card’ requirements”, and to allow “more creative artificial intelligence (AI)-enabled interoperability solutions”. The final rule of 28 October carries the same two sentences with “super intelligence” substituted and a footnote to EO 14434 attached. The proposal uses “artificial intelligence” 20 times; the final rule uses it once, in the parenthesis recording the chief AI officer’s old title. Nothing in the regulatory text turned on the word. What turned was the decision to delete (b)(11)(iv) to (vi), proposed under the old name and finalised under the new one.

The label changes no duty. The instruments that carry it do.

The wider record shows an executive branch still writing in both languages. A full-text search of the Federal Register, run by AI & GPP on 2 November for documents published since 29 September, returns 36 documents matching “artificial intelligence” and 12 matching “super intelligence”; seven are on both lists, because they quote statute, commenters, earlier rules or older titles. The helium notice says “SI infrastructure” in one sentence and cites “Winning the AI Race: America’s AI Action Plan” in the next, preserving a published title, as the State cable’s exceptions also do. The National Institute of Standards and Technology’s (NIST) announcement of 16 October takes the opposite course: it calls the same document “Winning the Race: America’s SI Action Plan” and quotes its recommendations with “AI” replaced by “[SI]” in square brackets, while citing NIST’s authority under the “National Artificial Intelligence Initiative Act of 2020”. The key practices themselves (PP-26-4396, version 1.0) are “ten voluntary key practices” for “the frontier SI industry”, examined in last week’s cover story on agent rulebooks.

Agencies are also rewriting what they have already published. NASA’s super intelligence page, last updated on 6 October, says the agency “is working to update its communications to incorporate the term ‘super intelligence’ as directed” and that its engineers “have safely used super intelligence for decades”. A NASA learning article for pupils in grades five to eight, updated on 27 October 2026, now opens: “Super intelligence, or SI, is a type of technology that helps machines and computers have ‘thinking’ abilities similar to humans.” Space.com reported on 16 October, from an internal plan it had seen, that NASA is working through four phases between 9 October and 23 December covering websites, staff titles and internal IT, that one platform’s URL change is described internally as a “breaking change for APIs and other integrations”, and that guidance for scientific publications is still being written. That account rests on Space.com alone; NASA’s public pages confirm the direction but not the schedule.

The pattern matches what AI & GPP found when the IMF produced a third term for the same technology: institutions change the noun and keep the definition, and the cost lands on whoever maintains the cross-references.

What the 28 November proposal could change, and for whom

Until Congress acts, the only legal content of “Super Intelligence” is a pointer to 15 U.S.C. 9401(3). Section 3(b) of the order asks the science adviser for something different: a definition that “reflects the capabilities described in section 1”. If the proposal narrows the term to frontier capability, every federal instrument that now says “SI” where it once said “AI” would, on its face, cover less than it did; if it widens the term, those instruments widen with it. The Treasury proposal is the exception, because it writes its own definition in section 851.232 and asks for comment on it. The order’s own answer is section 3(a): the pointer definition governs “unless and until superseded” by presidential action or an Act of Congress.

Three groups should read the November proposal when, or if, it appears. Buyers whose contracts tie the technology to a published definition, including those placed under the federal large-language-model clause that took effect on 19 October, are protected for now by section 2(b) and by the State cable’s exclusion of contracts and grants; a conforming amendment to 15 U.S.C. 9401 would reach any clause that cites that section. Investors with China exposure have until 7 December to argue about section 851.232 and the absence of any threshold, on which Treasury “particularly welcomes” comment. Health systems and their vendors have until 28 December before the model-card duty lapses, after which, in ONC’s words, information about intended use, training data and known risks is not “unknown or unavailable” but must be asked for.

What decision-makers should do now

The practical test for a ministry, a regulator or a vendor outside the United States is a drafting one. Open the contract, the risk register and the incident-reporting duty and find the defined term. If it is “artificial intelligence” tied to a published definition, American source material labelled “SI” still falls inside it, because EO 14434 says so; a document register that searches only for “AI” will nonetheless start missing it, a point made in last week’s Briefing. If the term is “super intelligence” borrowed from a federal template, the clause currently means the statutory definition and nothing more, and should say so until Congress gives the words content of their own. Three dates fix the calendar: 25 November for the Coast Guard docket, 28 November for the science adviser’s proposal and 7 December for Treasury. The fourth, 28 December, is the day a transparency requirement for clinical AI disappears, under a new name, for reasons that have nothing to do with the name.

Questions this answers

Does Executive Order 14434 change the legal definition of AI in the United States?
No. Section 3(a) of the order defines “Super Intelligence” and “SI” as the technologies encompassed by the statutory definition of artificial intelligence in 15 U.S.C. 9401(3), and section 2(b) requires no alteration of existing regulations, contracts or grants. A proposed statutory definition is due from the Assistant to the President for Science and Technology within 60 days, by 28 November 2026; only Congress can enact it.

What does the ONC final rule remove, and when?
The rule of 28 October 2026 removes 45 CFR 170.315(b)(11)(iv), (v) and (vi), the source-attribute “model card” support, access and modification requirements and the intervention risk-management requirement for predictive decision-support tools, together with the maintenance duty in 170.402(b)(4). The removals take effect on 28 December 2026. ONC says it may consider restoring elements of source-attribute transparency in future rulemaking.

What would the Treasury proposal prohibit, and by when must investors respond?
The proposed 31 CFR Part 851 would make the development of any “SI system” by a covered foreign person in a country of concern a prohibited technology, with no compute or benchmark threshold, replacing the tiered treatment of AI under the Part 850 rules in force since January 2025. Written comments must reach Treasury by 7 December 2026.

Sources

1. Executive Order 14434 of September 29, 2026, Inaugurating the Era of Super Intelligence — Federal Register via GovInfo, 2 October 2026
2. 15 U.S. Code § 9401 — Definitions — Legal Information Institute, Cornell Law School, accessed 2 November 2026
3. US diplomats told to say ‘super intelligence’ — not ‘artificial intelligence’ — after Trump’s call — The Associated Press, 23 September 2026
4. State Department orders ‘super intelligence’ wording across all official communications — Nextgov/FCW, 20 October 2026
5. Executive Order 14438 of October 22, 2026, Decreasing Housing Costs for the American People — Federal Register via GovInfo, 28 October 2026
6. Health Data, Technology, and Interoperability: ONC Deregulatory Actions To Unleash Prosperity; Final Rule — Federal Register via GovInfo, 28 October 2026
7. Health Data, Technology, and Interoperability: ASTP/ONC Deregulatory Actions To Unleash Prosperity; Proposed Rule — Federal Register via GovInfo, 29 December 2025
8. Outbound Investment Security Program Regulations; Proposed Rule — Federal Register via GovInfo, 26 October 2026
9. Treasury Issues Proposed Rule to Implement Comprehensive Outbound Investment National Security Act — U.S. Department of the Treasury, 22 October 2026
10. US Treasury Proposes Sweeping Limits on US Investments — Sidley Austin, 28 October 2026
11. Critical Material List; Addition of Helium — Federal Register via GovInfo, 27 October 2026
12. Cooperative Research and Development Agreement: Super Intelligence (SI) Situational Awareness Tools for Analyzing Coast Guard (CG) Sensor Feeds for Search and Rescue (SAR) — Federal Register via GovInfo, 26 October 2026
13. Accreditation, Innovation, and Modernization: The Secretary’s Recognition of Accrediting Agencies; Final Rule — Federal Register via GovInfo, 30 October 2026
14. Key Practices for Incident Readiness and Response in Super Intelligence Training and Evaluation, PP-26-4396, v1.0 — NIST CAISSI, NSA SISC and CISA, October 2026 (PDF)
15. Key Practices for Incident Readiness and Response in Super Intelligence Training and Evaluation — NIST news, 16 October 2026
16. Super Intelligence — NASA, page last updated 6 October 2026
17. What is Super Intelligence? (Grades 5-8) — NASA, 1 December 2025, page last updated 27 October 2026
18. ‘We are getting a lot of pressure from the administrator himself’: How NASA is racing to obey Trump’s ‘Super Intelligence’ order — Space.com, 16 October 2026
19. What’s in a Name? Executive Order Directs Agencies to Say “Super Intelligence,” Not “Artificial Intelligence” — Arnold & Porter, 13 October 2026