Super Intelligence vs AI: the new federal name uses the 2020 definition of AI
On 29 September 2026, President Trump signed Executive Order 14434, "Inaugurating the Era of Super Intelligence." It tells every federal department and agency to write "Super Intelligence" and "SI" where they used to write "Artificial Intelligence" and "AI." Within a week, California ordered the opposite, Elon Musk said SpaceXAI would become SpaceXSI, and a lot of people started asking whether AI had turned into something new.
We read the order in full. The line that answers that question is in Section 3(a), and few headlines quote it. For the purposes of the order, "Super Intelligence" and "SI" mean the technologies defined in section 9401(3) of title 15 of the US Code. That section is the definition of artificial intelligence written into the National Artificial Intelligence Initiative Act of 2020.
So the federal government has put a new name on the same legal definition. Below we cover three things: what the order changes, what "superintelligence" meant before last week, and what to do about it if you build or buy software.
The facts in five lines
- Executive Order 14434, signed 29 September 2026, replaces "AI" with "SI" in federal correspondence, websites, reports and policy documents.
- SI means the technologies covered by the 2020 federal definition of AI.
- Regulations, contracts and grants issued before the order stay as they are.
- A proposed legal definition of SI is due within 60 days, which is 28 November 2026.
- California's state agencies keep saying AI, under Executive Order N-10-26 of 30 September 2026.
1. What the order changes, and where it stops
Section 1 gives the reason for the change. The order says today's technologies go beyond what "Artificial Intelligence" describes and mark "a new era of Super Intelligence." It also says the modern field of AI was born in the United States, "where American scientists first gave the field its name," so the new name is presented as the next one.
Section 2 lists where the new words apply: "official correspondence, public communications, websites, reports, policy documents, and other non-statutory documents." Non-statutory means anything other than the text of a law. Section 2(b) sets the limit: nothing in the order requires changes to "previously issued regulations, Presidential actions, contracts, grants, or other historical documents." A federal contract signed in August that says "AI" stays exactly as written.
Section 3(b) starts the part that could matter later. The Assistant to the President for Science and Technology has 60 days to send proposed legislative language: a federal definition of Super Intelligence, an assessment of whether it should replace the existing legal definitions of AI, and the edits other laws would need. Only Congress can change a definition written into law. Until it does, SI in a federal document means what AI meant in the 2020 Act.
Here is that definition, read literally. Artificial intelligence is "a machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations or decisions influencing real or virtual environments." That wording covers a chatbot, a fraud score, and the system that recommends the next video. Under the order, all of them are now SI in federal writing.
Jennifer Dickey, a data privacy and AI attorney at Dykema, told the IAPP that the practical problem is managing "two vocabularies that may refer to the same systems in some settings but not necessarily in every legal or operational context." That is the risk worth planning for, and we come back to it in section 5.
2. What "superintelligence" meant before last week
The word has a long history in AI research, and there it means something specific. Nick Bostrom's 1997 paper "How Long Before Superintelligence?" 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 Future of Life Institute uses "artificial intelligence capable of outperforming all humans at most cognitive tasks."
The large AI companies started using the word in that sense over the past two years, and they describe it as a goal ahead of them. Here is the record, each item from the company or a named report:
| Date | Who | What they said or did |
|---|---|---|
| January 2025 | Sam Altman, OpenAI | "We are now confident we know how to build AGI as we have traditionally understood it. We are beginning to turn our aim beyond that, to superintelligence in the true sense of the word." |
| 30 June 2025 | Meta | Created Meta Superintelligence Labs to lead its AI research. |
| 30 July 2025 | Mark Zuckerberg, Meta | "Developing superintelligence is now in sight." The letter's goal is "personal superintelligence" for everyone. |
| August 2025 | Sam Altman, on CNBC | Called AGI "not a super-useful term." |
| 6 November 2025 | Mustafa Suleyman, Microsoft AI | Announced the MAI Superintelligence Team and its goal of "humanist superintelligence." |
| July 2026 | Alexandre LeBrun, AMI Labs | "Nobody is using it anymore; they switched to superintelligence." |
| 29 September 2026 | The White House | Executive Order 14434 makes SI the federal word for AI. |
So before the order, the industry had already moved from AGI (artificial general intelligence, a system roughly as capable as a person across tasks) to superintelligence (a system beyond any person). The order took the second word and attached it to the broad 2020 definition, which includes systems that existed long before either word was popular.
3. Why the name keeps changing
The people involved give their own reasons, and they differ.
The old word stopped meaning one thing. Altman's complaint about AGI in August 2025 was that it had stopped being useful. LeBrun, whose company avoids both terms, said the same of the replacement: "There's no good definition. What is superintelligence? I don't know. It's not a very useful word." He also predicted the next step: "Next time we'll switch to something else."
A larger word describes a larger goal. When Altman wrote that OpenAI was turning its aim "beyond" AGI, and when Zuckerberg wrote that superintelligence was "in sight," each was telling readers that the goal was now larger than AGI. In both cases the word names something the company says it is building and has not yet built.
The government says the old word is too small. The order's stated reason is that current technology exceeds what "Artificial Intelligence" describes. The definition it chose, however, is the one written in 2020 for the systems of 2020.
Those three reasons explain the result. The same word now carries two meanings:
| Superintelligence in research | SI in federal documents | |
|---|---|---|
| Defined by | Bostrom 1997; Future of Life Institute | 15 U.S.C. 9401(3), from the 2020 Act |
| Covers | A system beyond the best humans in almost every field | Any machine-based system that makes predictions, recommendations or decisions for goals a person sets |
| Exists today? | Described by Meta and OpenAI as a goal | Yes, and has since before 2020 |
The difference has consequences. The Future of Life Institute's Statement on Superintelligence calls for "a prohibition on the development of superintelligence," and it showed 76,845 signatures when we read it on 6 October 2026. Those signers meant the research sense of the word. Read with the federal meaning, the same sentence would ask for a ban on recommendation systems and fraud scores.
4. Who is using the new name
SpaceXAI says it will. Asked on X whether SpaceXAI would be renamed SpaceXSI, Musk answered "Yes, we will make that change" and "No more AI. SI, it's better," Fox Business reported on 4 October 2026. He gave no date for the change.
California says it will not. Executive Order N-10-26, signed by Governor Gavin Newsom on 30 September 2026, orders every agency and department under his authority to keep saying "Artificial Intelligence" and "AI," "notwithstanding any rebranded or different terminology used by the federal government, unless inconsistent with the law."
The leading labs had not moved as of 1 October. The San Francisco Standard reported that reactions in Silicon Valley were mixed and that the leading labs had not relabelled their models as SI.
So a company selling to both the federal government and the State of California now meets two required vocabularies for the same product.
5. What to do if you build or buy software
Three rules we would follow today.
Tie the word to a definition in every contract. A label on its own now means different things to different readers. Write the term with a fixed definition, for example "artificial intelligence or Super Intelligence, as defined in 15 U.S.C. 9401(3)," so the contract means one thing until the law changes. If you respond to federal solicitations, expect SI in new ones and read it as AI until Congress says otherwise. Have your own lawyer check the wording.
Wait for the definition before renaming anything. The proposed legal definition is due on 28 November 2026, and it is a proposal for Congress, which has not acted. If you rename your product, your documentation and your sales material now, and the final definition differs, you pay for the rename twice.
Ask which meaning a vendor intends, then ask for results. When a vendor calls its product superintelligent, ask whether they mean the federal sense (it makes predictions or decisions) or the research sense (it is beyond the best humans). The first is true of almost every AI product on the market. The second describes what Meta and OpenAI call a goal they are still working toward. Then ask for test results on your own tasks, with your own data, because neither definition tells you whether the system does your work correctly.
That last rule is how we work regardless of the name. We write the eval suite, the automated checks a build has to pass, from the specification before the code exists. Each check feeds the system a known input and compares its output to the expected answer. A change of label in federal documents leaves every one of those checks exactly as it was.
Credit for this piece goes to the reporters at the IAPP, TechCrunch, Fortune, Fox Business and the San Francisco Standard, whose work we checked each claim against, and to the White House and the California Governor's office for publishing both orders in full.
A name tells you what someone wants you to believe about a system. A test tells you what the system does.
Sources
- The White House, Executive Order 14434, Inaugurating the Era of Super Intelligence, 29 September 2026
- Cornell Legal Information Institute, 15 U.S. Code § 9401, Definitions
- State of California, Executive Order N-10-26, 30 September 2026
- IAPP, AI by any other name? White House directs federal agencies to enter the era of 'super intelligence'
- San Francisco Standard, Trump wants to rebrand AI "Super intelligence." Will Silicon Valley play along?, 1 October 2026
- Fox Business, Elon Musk renames SpaceXAI to SpaceXSI, 4 October 2026
- TechCrunch, Kate Park, Why AMI Labs' Alexandre LeBrun won't call his AI 'AGI' or 'superintelligence', 16 July 2026
- Nick Bostrom, How Long Before Superintelligence?, 1997
- Sam Altman, Reflections, January 2025
- Fortune, Sharon Goldman, AGI talk is out in Silicon Valley's latest vibe shift, 25 August 2025
- Fortune, Key takeaways from Mark Zuckerberg's memo on Meta's new superintelligence effort, 2 July 2025
- Meta, Personal Superintelligence, 30 July 2025
- Microsoft AI, Mustafa Suleyman, Towards Humanist Superintelligence, 6 November 2025
- Future of Life Institute, Statement on Superintelligence, read 6 October 2026, and press release
