Super Intelligence Executive Order: What It Actually Changes
The Super Intelligence executive order explained: what Trump’s AI-to-SI order requires of federal agencies, what it leaves alone, and the 60-day deadline.
The short answer: the Super Intelligence executive order changes a word, not a technology. On Tuesday 29 September 2026 President Trump signed Inaugurating the Era of Super Intelligence, which tells federal agencies to write “Super Intelligence” and “SI” wherever they used to write “artificial intelligence” and “AI”. It applies to the executive branch. It does not tell businesses, schools or the public to do anything, and the tools you use at work are exactly what they were on Monday.
The name itself is not new. It first surfaced in a UN speech and then at the US-China summit the week before, which the US-China Super Intelligence Dialogue explainer covers. What is new is that it is now written policy, with a deadline attached. This post goes through what the order says, what it leaves alone, and the one part worth watching.
What the order requires
The order is short, four sections. Read against the White House fact sheet, it does three things:
- Swaps the vocabulary. “To the maximum extent permitted by law”, executive departments and agencies must use “Super Intelligence” and “SI” in official correspondence, public communications, websites, reports, policy documents and other non-statutory documents.
- Keeps the meaning. The new terms refer to the same technologies already covered by the federal definition of artificial intelligence in 15 U.S.C. 9401(3), which describes a machine-based system that makes predictions, recommendations or decisions for human-defined objectives. In other words, SI legally means what AI meant yesterday.
- Starts a clock. Within 60 days, the Assistant to the President for Science and Technology must send proposed legislative language for a federal statutory definition of “Super Intelligence”, including whether it should modify, expand upon or supersede the existing definitions of artificial intelligence in law.
The stated reason is in Section 1: frontier systems “do much more than imitate or automate” parts of human intelligence, so “artificial” undersells them. Trump put it more plainly at a White House event that day, as reported by Business Insider: “We don’t want to hear artificial. Because it’s not artificial.”
What the order does not change
This is the part most headlines skip, and it is most of the order’s effect.
- Old paperwork stays as it is. Section 2(b) says nothing in it requires rewriting previously issued regulations, presidential actions, contracts, grants or other historical documents. A federal contract that says “AI” still says “AI”.
- Laws stay as they are. An executive order cannot amend a statute, which is why the order asks for proposed legislation rather than changing the definition itself. Until Congress acts, US law keeps saying “artificial intelligence”.
- Private organisations are not addressed. The text speaks to executive departments and agencies. It says nothing about companies, states or individuals, and it creates no right or benefit “enforceable at law or in equity”, which is standard wording for orders of this kind.
- The technology does not change. No model gets more capable because a government calls it something new.
So if you work for a company that sells to the US government, expect to see “SI” in new solicitations and press material. It will almost always mean ordinary AI. Nothing in the order tells you to change your own documents.
SI is not superintelligence in the research sense
This is where the name causes real confusion. Among researchers, superintelligence means a hypothetical system that far outperforms people at almost every intellectual task. OpenAI described it that way in its 2023 post on governance of superintelligence, as something that does not exist yet and would need new kinds of oversight if it ever did.
The order’s “Super Intelligence” is a different thing: a new label for the broad set of systems the old federal AI definition already covered, including the chatbots people use today. Axios pointed out the mismatch the week before the order, when the name was first used at the US-China summit. Same words, two meanings. When you read “SI” in a US government document, translate it as “AI”. It is the same kind of naming trap as the one in What Does GPT Stand For, where the letters tell you something useful only once you know what they refer to.
The practical risk is expectation. A tool described as “super intelligent” still invents sources and is confidently wrong in exactly the ways described in What AI Is Actually Bad At. The new name does not come with a new error rate.
The accord signed the same day
The order landed on the same day as a separate document, and the two are easy to mix up. Leaders of Anthropic, OpenAI, Google, Meta, xAI and Nvidia signed a voluntary pact at the White House, reported by Al Jazeera as the Joint Commitment on Frontier Responsibilities. ABC News gives its full title as “The White House Accord on Superintelligence: A Joint Commitment on Frontier SI Responsibilities”.
According to Nextgov, it sets out four layers: internal controls to monitor model capabilities, an internal team to check those controls work, an independent external auditor, and board-level oversight. It is not law. Trump called it “morally binding”, per ABC, and The Register notes it has no enforcement mechanism and only says it may make sense to codify the steps later. Critics quoted by Al Jazeera ask why an industry should mark its own homework. On the other side, Vice President JD Vance argued, per Nextgov, that officials know far less about the technology than the people building it. Both positions are on the record.
For someone using these tools at work, the accord changes nothing you can see today. What happens to the text you type into a chatbot still depends on that product’s settings and your employer’s rules, which is the question Is GPT-6 Sol Safe works through for one current model.
What to watch next
The 60-day definition is the only part of the order with legal weight still to come. Counting from 29 September, the proposal is due by about 28 November 2026. Three things will tell you whether it matters:
- Whether the proposed definition simply renames the existing one, or draws a new line around a narrower class of systems. A narrower one would change which rules apply to which tools.
- Whether Congress takes it up. A proposal sent to Congress is not a law until it passes.
- Whether the SI label spreads beyond the federal government. So far it has not: Beijing’s own account of last week’s summit still said “artificial intelligence”.
What this means for you
If you use AI at work, carry on. Write “AI” in your own documents if that is what your organisation uses; the order does not reach you. If you deal with US federal agencies, read “SI” as “AI” unless a document defines it otherwise, and ask if it is unclear. If a vendor starts selling a product as “super intelligent”, ask what specific task it does reliably and test it on your own examples, the way How to Check an AI Answer When You Are Not the Expert describes. And if your team is choosing tools, start from the job to be done rather than the label, as in Best AI Tools for Business.
Whatever Washington calls it, the useful skill has not changed: give the tool a clear task and check what comes back. That is what Coursium teaches, in short lessons on your phone. Its certificate records that you finished a program; it is not an accredited qualification.