Evidence as of October 4, 2026. Every statement below comes from a retained, dated source in this story's evidence packet, or from the video record page. The packet records dates to the day only, and is explicit about its limits: attribution, source identity and located excerpts are checked, while "[t]ruth, rights clearance and public release require editorial review."
On September 29, 2026 the President signed Executive Order 14434, "Inaugurating the Era of Super Intelligence." The same afternoon six technology executives signed a voluntary accord at the White House. On October 4 a "Super Intelligence Force" was announced on Truth Social. The instruments that would compel anyone — bills — sat in committee without a vote.
Watch the video record. The video carries the original recordings with their own audio; this is the written companion.
The order renames; it does not legislate
Section 2(a) directs agencies, "[t]o the maximum extent permitted by law," to use "Super Intelligence" and "SI" in place of "Artificial Intelligence" and "AI" in "official correspondence, public communications, websites, reports, policy documents, and other non-statutory documents." Section 2(b) requires no alteration of regulations, presidential actions, contracts or grants already issued.
The definition is the part most easily missed. Section 3(a) defines the new terms as the technologies already "encompassed by the term 'artificial intelligence' as defined in section 9401(3) of title 15, United States Code" — "a machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations or decisions influencing real or virtual environments." The order substitutes a label, not a capability threshold. Section 3(b) gives the President's science adviser 60 days — to November 28 — to propose legislative language for a federal definition. Section 4 adds that it creates no right "enforceable at law or in equity."
On September 30 the IAPP's Jennifer Dickey (Dykema) expected the result: "SI" in federal materials while statutes, state laws, contracts and vendor questionnaires keep saying "AI."
The accord: four layers, no penalty
Its subtitle is "Joint Commitment on Frontier Responsibilities" (Washington Examiner published the full text). Al Jazeera's September 30 explainer describes four layers: internal controls; an internal team to oversee them; independent outside auditors; and an independent board committee to review the results. The same report states what it does not do: it is voluntary and not legally binding, sets no penalties, requires no public disclosure of audit results, and gives the government no enforcement role. Its own text allows that over time it might "make sense to codify these steps into laws or regulations." The President called it "morally binding" and promised "a tremendous self-policing aspect."
Al Jazeera lists the signatories as Meta, Nvidia, Google, OpenAI, xAI and Anthropic; Mashable notes that "neither Microsoft nor Amazon signed the accord." Medill on the Hill lists SpaceX where Al Jazeera has xAI; we record the discrepancy rather than resolve it.
TechCrunch, citing the Wall Street Journal, reports Director of National Intelligence Jay Clayton chairing the new Force, a report on risks and opportunities due in 120 days, and a charter to plan for "SI-enabled threats to our society, while preventing overregulation and regulatory capture." Clayton told the Journal that "the risk of not being first is high."
Five positions on one question
The question in the record is who, if anyone, slows the technology down. The labels are Comuvia's arrangement of the record, not claims any party made.
| Position (Comuvia's label) | Who holds it | Dated evidence |
|---|---|---|
| Companies audit themselves | The President and the six signatories | EO 14434 and the accord, Sept. 29 |
| Coordinate, but do not slow down | The Super Intelligence Force | TechCrunch, SiliconANGLE, Oct. 4 |
| Disclosure by law, enforced by the FTC | Senators Coons, Britt, Schatz, Lankford | Coons release, Sept. 24 |
| Ban, pause, and a new department | Senator Sanders, Rep. Casar | Sanders, Casar releases, Sept. 23 |
| Outside limits on the pace itself | Dario Amodei; the GovAI authors | Amodei essay, Sept. 12; GovAI, Sept. 28 |
The AI Systems Transparency Act, announced on September 24, would have the Federal Trade Commission enforce disclosure of the material usually found on "model cards"; Senator Coons said companies "must not be allowed to hand wave away these doubts." The Ban Artificial Superintelligence Act, per Senator Sanders's September 23 release, would create a cabinet-level Department of Artificial Intelligence and set penalties of "not more than 20 years in prison" for individuals. GovTrack shows that bill, S. 5493, referred to Senate Commerce on September 23 with a "5% chance of getting past committee." As of October 4 the packet records no committee vote on any of the three bills.
The dispute, both sides
Gary Marcus, professor emeritus at New York University, wrote on September 29 that the accord read as "We agree not be regulated," and told PBS NewsHour that evening that self-regulation "is certainly not enough," asking for outside regulation "just like we have in aviation and food and drugs." The Council on Foreign Relations' Connor Martin wrote on October 1 that "there is nothing in the document that legally compels any of the signers to change their behavior." At a Senate subcommittee hearing on September 30, Senator Richard Blumenthal called the scheme "completely voluntary and totally secret" and "worse than ineffectual."
The other side does not deny that the technology is serious. Sundar Pichai framed the accord as a commitment "to establish norms and build public confidence," and OpenAI's Chris Lehane had said industry standards would "complement — not replace — mandatory federal safeguards." On September 12 Dario Amodei argued that "[w]e must slow the pace at which we improve the capabilities of AI models," through "regulation that targets all US frontier AI companies." On September 28, 22 authors including Geoffrey Hinton and Yoshua Bengio wrote that an intelligence explosion "would compress years of progress into months," warning, as The Guardian reported, that "[o]nce an intelligence explosion begins, the window for action may close."
What the rest of the record shows
Nobody else changed the word. Norton Rose Fulbright reported on July 24 that the EU's Digital Omnibus on AI had moved high-risk obligations to 2 December 2027 and 2 August 2028. In the UK, a private members' bill from Labour MP Alex Sobel would curb systems able to "render ineffective relevant human authorities"; The Next Web reported that the government does "not consider the measures proposed in the bill to be the right approach." MMLC Group reported in August that "China has not yet enacted a dedicated" AI law. UN News quoted Yoshua Bengio on July 5 that "science currently cannot guarantee that as capabilities continue to increase, AI will not cause catastrophic harm."
Public attitudes were already cooling. Gallup's July 28 release (3,270 US adults, fielded May 4–11, 2026) found 27% trusting businesses to use AI responsibly, down from 31%; 39% saying AI does more harm than good, up from 31%; and 79% expecting fewer US jobs over ten years, up from 73%.
Two mechanisms decide how this ends. An executive order can be undone by a later one: Executive Order 14179 of January 23, 2025 revoked its 2023 predecessor that way. And a bill that does not pass dies with the Congress — the Senate's reference states that "Congresses have begun and ended at noon on January 3 of odd-numbered years."
Comuvia's judgement, labelled
This paragraph is a Comuvia judgement, not a finding from the record. We rate the complexity high, unchanged from our September assessment. The facts are clean — documents and recordings, all on the record — while the dispute is about who decides. Four kinds of instrument are in play: a renaming order, an accord with audits but no penalty, a task force whose charter is a post, and bills with no committee vote. Observability is the weak point: if audit results need not be published, a reader cannot tell compliance from non-compliance. We hold no position on which is correct. Nothing here is a view on any company's prospects or on policy costs. Not investment advice.
A note on form: the video companion uses AI-generated narration and a synthetic presenter, and says so on screen. The record fragments inside it are the original recordings, credited individually; the full source list with dates is on the record page.