Super Intelligence Executive Order: What Actually Changes

On September 29, 2026, President Trump signed an executive order directing the federal government to replace "artificial intelligence" and "AI" with "Super Intelligence" and "SI," according to the White House fact sheet. Within a week, a new White House task force and a corporate rename followed. This piece separates what the Super Intelligence executive order changes in law from what it changes only in language, and looks at what that means for procurement, compliance and anyone who sells to government.

Sep 29, 2026Order signedExecutive order plus industry accord

60 daysDefinition deadlineProposed legislative language on SI

6Accord signatory companiesGoogle, Anthropic, Meta, OpenAI, xAI, NVIDIA

4Control layersInternal, team, external audit, committee

71 yearsSince Dartmouth1955 proposal to 2026 order

What the Super Intelligence executive order says

Per the White House fact sheet, the order applies across federal departments and agencies and covers official correspondence, public communications and policy documents. It excludes statutory regulations and contracts. It also instructs the Assistant to the President for Science and Technology to propose a federal definition of "Super Intelligence" and to identify further executive actions. The stated rationale is that the technology is "unlocking new forms of human creativity and empowering Americans to achieve what was previously impossible, not simply imitating or automating human intelligence."

Nextgov reports that the science adviser has 60 days to propose legislative language defining SI, including whether it should modify or supersede existing AI definitions. GovConWire quotes the operative standard: agencies adopt the new terms in "official communications, websites, reports, policy documents and other non-statutory materials, to the maximum extent the law allows," and historical documents, contracts and grants are exempt. Counting 60 days from signing lands around November 28, 2026; that date is our arithmetic, not a published deadline.

What legally changes, and what does not

The order changes vocabulary in agency-authored material. It does not rewrite the statute book. The binding federal definition of artificial intelligence sits in 15 U.S.C. 9401(3), enacted by the National Artificial Intelligence Initiative Act of 2020 (Pub. L. 116-283), which defines it as "a machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations or decisions influencing real or virtual environments." An executive order cannot amend that text. According to GovConWire, the legal definition remains unchanged pending further action, and the 60-day proposal is the route to changing it.

The order changes what agencies write. The statute still says what the law means.
Metir analysis of the order's exemptions

In practical terms, the picture breaks into three layers:

Changes now: agency websites, press releases, reports and policy documents, as far as the law allows.
Unchanged for now: statutes, regulations, existing contracts and grants, per the fact sheet and GovConWire.
Open: whether Congress adopts a new definition, which depends on the proposal the science adviser delivers.

From Dartmouth to the Super Intelligence order

Key dates for the term, the statute and the 2026 order.

Sep 2, 1955
Dartmouth proposal
McCarthy, Minsky, Rochester and Shannon propose a summer study of "artificial intelligence".
Jun to Aug 1956
Dartmouth workshop
Run from about June 18 to August 17, per Ray Solomonoff's notes.
2014
Bostrom's Superintelligence
Oxford University Press publishes the book that defines the research-literature term.
Jan 1, 2021
Statutory definition
15 U.S.C. 9401(3) enacted through Pub. L. 116-283.
Sep 29, 2026
Executive order and accord
Agencies told to say "Super Intelligence"; statutes, regulations and contracts exempt.
Oct 4, 2026
Force named, SpaceXSI
Super Intelligence Force leaders announced; Musk says the SpaceXAI rename will happen.
About Nov 28, 2026
60-day mark
Deadline for proposed legislative language defining SI, counted from signing.

Where the term "artificial intelligence" came from

The phrase is a deliberate coinage. The Dartmouth workshop began with a proposal presented on September 2, 1955 by John McCarthy, Marvin Minsky, Nathaniel Rochester and Claude Shannon. It conjectured that "every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it." McCarthy chose the name partly for its neutrality, avoiding a focus on narrow automata theory and avoiding cybernetics. The summer project ran from about June 18 to August 17, 1956, according to Ray Solomonoff's notes cited in that article.

Dartmouth Hall, a white Georgian-style building with black shutters on the Dartmouth College campus in Hanover, New Hampshire — Dartmouth Hall on the Dartmouth College campus in Hanover, New Hampshire, photographed in 2007. The 1956 workshop that popularized the term "artificial intelligence" was held at Dartmouth College; this photo shows the campus, not the workshop. Photo: Kane5187, public domain.

So the 2026 order is not the first time the field's name has been a policy choice. It is, however, a rare case of a government mandating a replacement for it.

"Superintelligence" in the research literature

In research usage the word carries a specific, much stronger meaning. Philosopher Nick Bostrom defines superintelligence as "any intellect that greatly exceeds the cognitive performance of humans in virtually all domains of interest," per the Wikipedia summary of the concept. His 2014 book, Superintelligence: Paths, Dangers, Strategies, was published by Oxford University Press. The same source describes three forms: speed superintelligence (human-like reasoning, much faster), collective superintelligence (many systems coordinating) and quality superintelligence (fundamentally better cognition).

Philosopher Nick Bostrom gesturing while speaking on stage in an armchair — Nick Bostrom speaking at Collision 2019, where he was introduced as Director of the Future of Humanity Institute at Oxford. Photo: Collision Conf, CC BY 2.0.

The order does not adopt that definition; it asks for a new federal one. The fact sheet's rationale frames SI as technology that goes beyond "imitating or automating human intelligence," which is a descriptive claim about capability rather than the research-literature threshold of exceeding humans in virtually all domains. That gap is why the pending definition matters. Until it exists, "SI" in an agency document is a label for the same class of systems the statute already covers as AI. The mismatch with research usage is an observation, not a judgment on the choice.

The accord and the Super Intelligence Force

The same day, executives signed a separate voluntary pact. Nextgov lists the signatories as Trump, Google's Sundar Pichai, Anthropic's Dario Amodei, Meta's Mark Zuckerberg, OpenAI's Greg Brockman, xAI's Elon Musk and Nvidia's Jensen Huang. The accord sets four layers: internal controls monitoring model capabilities and alignment, a dedicated internal oversight team, independent external auditors, and an independent committee reviewing audit reports. Fox Business reports that Trump called the document "morally binding," while GovConWire notes the signatories pledged implementation regardless of legal requirement.

Google

Anthropic

Practical implications for procurement, search and compliance

The following is analysis, not reporting. Because contracts, regulations and statutes are exempt, existing language such as "artificial intelligence" in solicitations, clauses and compliance frameworks stays valid for now. Over time, three effects are plausible:

Dual vocabulary in proposals. Agency-authored documents will say SI while statutes and legacy contracts say AI. Vendors who mirror only one term risk missing keyword filters on either side.
Search behavior shifts gradually. Buyers and analysts searching agency sites will encounter "Super Intelligence" in new pages, while older pages, rules and contract vehicles keep the old term. Content that names both terms, with plain definitions, covers both.
Compliance mapping needs a crosswalk. If the 60-day proposal alters the statutory definition, obligations keyed to "artificial intelligence" could be reinterpreted. Teams should map each internal control to the current statutory text and track the proposal rather than renaming controls early.

For teams building on this, keep terminology in configuration and documentation, not hard-wired into systems. A model-agnostic workspace like Metir makes that easier because the label on a provider or product can change without changing how work flows.

Bottom line

The order is a language directive with a defined follow-up: a proposed definition due in about two months. It leaves the 2020 statutory definition, regulations and contracts as they are, and it comes with a voluntary accord and a task force that, per reporting, has no statutory authority or budget. The rename is already spreading into corporate branding. Whether "Super Intelligence" becomes law, stays a house style, or fades is something the definitional proposal, and Congress, will decide.

Sources:

Image credits

Dartmouth Hall, Dartmouth College, Hanover NH (2007): photo by Kane5187, public domain. Wikimedia Commons file.
Nick Bostrom at Collision 2019: photo by Collision Conf, licensed CC BY 2.0. Wikimedia Commons file.