On September 30, 2026 the President signed an executive order telling every federal agency to say "super intelligence" (SI) instead of "artificial intelligence" (AI). The next day the largest labs signed a "White House Accord on Super Intelligence." The technology did not change. The word did, for the government. This page separates the two.
How this site handles it. Frontier // Decoded is written in "AI", the word 53% of readers still use. Flip the switch and every page reads in "SI" wording — your choice is remembered. The switch also lives in the header on every page.
Seventy years from a summer workshop to an executive order.
John McCarthy uses the phrase for the Dartmouth summer workshop that founded the field.
I. J. Good describes an "intelligence explosion"; Nick Bostrom's 2014 book Superintelligence makes the term mainstream.
Ilya Sutskever founds Safe Superintelligence Inc.; Meta renames its AI division Meta Superintelligence Labs.
Federal agencies are directed to use "super intelligence" and "SI" in new communications, websites and policy documents, "to the maximum extent permitted by law."
Leaders from Google, Anthropic, Meta, OpenAI, xAI and Nvidia sign a "morally binding" pledge: internal control teams, outside auditors, published safety policies. Apple is absent.
A presidential term is not a law. Here is the exact line.
"We can say the word super, because the other word is a fake word."President Trump, on why "artificial" had to go · Sept 30, 2026
"The technology has real risks."Dario Amodei, Anthropic, at the accord press conference · Oct 1, 2026
"Trump's word games are adding confusion to this critical — and already confused — conversation."David Krueger, AI professor
YouGov asked Americans what the technology should be called after the order.
The frontier labs, what they call it, and the accord.
| Lab | Base | Flagship line | Uses the word "superintelligence" | Signed the Oct 1 accord |
|---|---|---|---|---|
| OpenAI | San Francisco | GPT family | In its mission statements; still said "AI" at the press conference | Yes |
| Anthropic | San Francisco | Claude family | Rarely; frames the work as frontier AI safety | Yes |
| Google DeepMind | London / Mountain View | Gemini family | Talks of AGI; Pichai used "AI" language at the accord | Yes |
| Meta Superintelligence Labs | Menlo Park | Llama family | Yes — in its own name since June 2025 | Yes |
| xAI | Palo Alto / Memphis | Grok family | Musk corrected himself to "SI" on stage | Yes |
| Nvidia | Santa Clara | Blackwell → Rubin silicon | Yes — "super intelligence factories" | Yes |
| Safe Superintelligence Inc. | Palo Alto / Tel Aviv | No public product | Yes — founded on the word, June 2024 | Not listed |
| Apple | Cupertino | Apple Intelligence | No | Not present |
The live numbers behind this race — accelerators, clusters, capex — are on CPU // Decoded, the foundation-model log and training clusters.
The news agent researches this beat every 8 hours; the Research Desk files a sourced report every two weeks.
Agencies will write "SI" from now on. The statute, the NIST framework and every existing contract still read "AI". The only binding deadline is the 60-day legislative proposal. Watch for that text, for whether Congress takes it up, and for whether a single lab changes the name of a product.