Artificial intelligence has been called many things in its 70-year history — overhyped, revolutionary, dangerous, a bubble. As of last month, the United States government has a new official position: it shouldn’t be called artificial intelligence at all.
In an executive order that landed with remarkably little technical debate, the Trump administration directed every federal agency to replace the terms “Artificial Intelligence” and “AI” with “Super Intelligence” and “SI” across official communications, websites, and policy documents. The technology did not change. The models did not change. The name did — and the fight over that name tells you more about where this industry is heading than most product launches do.
The order that renamed a technology
The executive order’s language is worth reading closely, because it is unusually honest about its own purpose. Federal terminology, it says, should “reflect the transformative capabilities of these technologies and the limitless opportunities they create for the American people.” The term Super Intelligence, the order continues, “more appropriately captures the promise, potential, and rapidly advancing capabilities of these technologies.” Agencies are to use the new terms “to the maximum extent permitted by law” and to not recognize the use of the old ones.
The president was blunter in public. AI, he has argued, has a branding problem — the word “artificial” sounds fake, and the term has become associated with fear and distrust. On Truth Social, he escalated the campaign into something closer to a loyalty pledge, declaring that the White House considers anyone who keeps using “Artificial Intelligence” instead of the new term to be “THE ENEMY!”
Alongside the order came a White House Accord on Super Intelligence — a voluntary, self-described “morally binding” pledge in which AI companies commit to internal controls and safeguards. It has no enforcement mechanism. It is, in effect, a constitution with no police force, signed by the very companies it governs.
None of this can legally compel a private company or a citizen to change their vocabulary. But Washington doesn’t need a law to move an industry that sells to the government, fears regulation, and competes for the president’s favor. It just needs a signal.
Big Tech falls in line (mostly)
The signal was received. Salesforce CEO Marc Benioff announced that the company had rebranded its agent platform from AIForce to SIForce, declaring, “The era of Super Intelligence is here.” Elon Musk pledged to rename SpaceXAI to SpaceXSI and updated the division’s social accounts. Microsoft CEO Satya Nadella has adopted the capitalized “Super Intelligence” phrasing in his public posts on AI governance. NVIDIA CEO Jensen Huang attended the White House summit accompanying the order, where the chipmaker committed $1 billion toward scientific advancement.
The most interesting response came from the company with the most to lose from a rename. OpenAI CEO Sam Altman confirmed his company has “no plans” to become OpenSI. Yet OpenAI president Greg Brockman attended the signing summit and signed the Accord on Super Intelligence. That split — sign the pledge, keep the name — is probably the template for how the industry handles this: compliance theater where it is cheap, resistance where it is expensive. OpenAI’s name is worth billions. A Salesforce product line is not.
Meanwhile, federal agencies have begun swapping terminology across their web properties, and anyone who sells software or services to the government now faces a practical question: do your proposals say AI or SI? Increasingly, the answer is determined by which word the customer’s style guide demands.
The scientists’ objection: words mean things
The research community’s pushback has been swift, and it is not really about politics. It is about a definition.
“Superintelligence” is not a marketing synonym for “very good AI.” In the technical literature, it refers to a hypothetical intellect that exceeds the cognitive performance of the best human minds in practically every domain — the concept at the center of decades of debate about AI’s long-term future, most famously in Nick Bostrom’s 2014 book Superintelligence. Whatever you think of today’s large language models, they are not that. They are statistical systems that predict text, images, and actions, brilliantly and usefully, within limits that surprise their own creators in both directions.
Renaming today’s systems “super” does two kinds of damage, critics argue. First, it inflates the expectation gap: a technology sold as superhuman will be judged as a failure every time it hallucinates a citation or drives a workflow into a wall. Second, it burns the vocabulary the field will need later. If 2026’s chatbots are already “super intelligence,” what do we call the thing the word was invented to describe, if it ever arrives?
There is also a simpler objection, made by more than one commentator: a rename imposed from above, enforced by loyalty rhetoric, is not how technical language is supposed to evolve. Terms win because practitioners find them accurate. “Artificial intelligence” itself was a branding choice — John McCarthy coined it in 1955 partly to distance the new field from cybernetics — but it won in the marketplace of researchers, not by decree.
The other SI: synthetic, not super
Complicating everything is a rival camp arguing that if the initials SI are going to be used at all, they should stand for something nearly opposite: Synthetic Intelligence.
The argument runs on an analogy from consumer law. Lab-grown diamonds are chemically identical to mined ones, but the U.S. Federal Trade Commission requires them to be labeled synthetic — sellers cannot call them “super diamonds.” Today’s AI models, the argument goes, are similarly manufactured: built in server farms from human-created data, industrial synthesis rather than digital minds. Calling them “synthetic” strips away the mysticism while acknowledging the power of the tool. Calling them “super” does the reverse.
This framing has commercial adopters of its own — the app-building platform Famous.ai markets its system as Synthetic Intelligence and declares AI dead, a claim reviewers have filed under branding rather than science. But the underlying point has serious defenders: names shape regulation, liability, and public trust, and a name that describes what the thing is will age better than a name that describes how it should make you feel.
So there are now three SIs in circulation — the administration’s Super, the critics’ Synthetic, and the research literature’s superintelligence — plus the term everyone actually uses: AI.
Why this matters beyond the politics
It is tempting to file the whole episode under political theater and move on. That would be a mistake, for three practical reasons.
Language is procurement policy. Federal spending shapes vendor roadmaps. If RFPs, compliance frameworks, and agency style guides say SI, then thousands of companies will say SI in every document that touches Washington — and that vocabulary will leak outward into enterprise marketing, exactly as government acronyms always do.
Names set expectations, and expectations set budgets. Executives who bought “AI” tools and got a helpful-but-fallible assistant are recalibrating. Executives sold “super intelligence” will expect the fallibility to be a bug that gets patched, not a property of the technology. The disappointment cycle that follows inflated names is how industries lose credibility they spent years earning.
The fight itself is information. The speed with which major CEOs adopted the new term — and the precision with which OpenAI declined to — is a map of who needs the administration’s goodwill and for what: government contracts, merger approvals, export licenses, energy for data centers. Watch the vocabulary and you can see the industry’s dependencies in real time.
So what should you call it?
For anyone building, buying, or writing about this technology outside the federal bubble, the answer is boring and clear: keep saying AI. It is the term your customers search for, your employees understand, and your contracts already use. Renaming your product SI in 2026 buys you no clarity and costs you recognition — with one exception. If you sell to the U.S. government, mirror your customer’s language in their documents, the same way you’d mirror any client’s house style.
And keep an eye on the synthetic camp. If any challenger term has a real future, it is the one that makes the technology sound less magical, not more — because the long-run trust problem for AI was never that people think it is fake. It is that people were promised it was super.
The name on the label will keep changing. The technology underneath will keep doing what it does: impressive, fallible, and entirely indifferent to what Washington calls it.
