Aktiver — semantic AI platform Deterministic · traceable · behind your firewall
Insights / Government

A new name for AI, and a voluntary accord on frontier model controls

On 29 September 2026 President Trump signed Executive Order 14434, “Inaugurating the Era of Super Intelligence,” directing federal agencies to use “Super Intelligence” and “SI” in place of “Artificial Intelligence” and “AI” in their non-statutory communications. Hours earlier, at a White House meeting on AI safety, leaders of six technology companies signed a voluntary accord on controls and audits for frontier models, Nextgov/FCW reported.

The two documents together set out the administration’s current approach: change the language, keep the legal definitions for now, and rely on industry self-governance rather than a new regulator.

/ 01What the order does

The text of the order states that “the terminology used by the Federal Government should reflect the transformative capabilities of these technologies.” Section 2 requires agencies to use the new terms in official correspondence, public communications, websites, reports and policy documents, to the maximum extent permitted by law. It explicitly does not require existing regulations, contracts, grants or historical documents to be changed.

Section 3 matters most for anyone who contracts with the government. For now, “Super Intelligence” means exactly what “artificial intelligence” means in the existing statutory definition at 15 U.S.C. § 9401(3). Within 60 days, by 28 November, the Assistant to the President for Science and Technology must propose legislative language: whether the new definition should modify, expand or supersede the statutory one, conforming amendments to statutory references, and any further executive action needed. The order creates no enforceable rights.

/ 02What the accord commits companies to

The accord was signed by the President and by Sundar Pichai of Google, Dario Amodei of Anthropic, Mark Zuckerberg of Meta, Greg Brockman of OpenAI, Elon Musk of xAI and Jensen Huang of NVIDIA. According to Nextgov/FCW, participating companies would maintain four layers of controls:

  • internal controls to monitor model capabilities and alignment during training and deployment, including cybersecurity;
  • an internal team to oversee those controls;
  • partnerships with independent external auditors;
  • an independent committee to review reports from the internal teams and external auditors.

The accord states that AI safety “starts with every company that is training and deploying frontier models having robust internal processes and controls.” Vice President JD Vance argued against an FDA- or FAA-style regulator for frontier models, saying the FTC and Justice Department already have authority to act on consumer harm.

/ 03Why it matters

The renaming will get the attention, but for practitioners it changes little in the short term: contracts, regulations and definitions stand. The 28 November deadline is the part to watch. A new statutory definition could widen or narrow what counts as AI in procurement, reporting and oversight rules, and that would ripple through every agency inventory and every vendor’s compliance paperwork.

The accord is more substantive. It is voluntary and limited to frontier developers, but its structure, internal controls, an oversight team, external audit and an independent review committee, is a recognisable assurance model. It shifts the question from whether a model is safe to whether a company can show evidence that it checked.

/ 04Labels change; evidence does not

For organisations deploying AI, by any name, the lesson is that terminology is unstable and definitions are about to be renegotiated. Systems and documentation that depend on a particular label will need rework. Systems that describe what a tool actually does, what data it uses, what decisions it informs and how its outputs can be checked, will not.

That is an argument for keeping the meaning of things in a formal model rather than in prose. In an ontology, a renamed concept is a new label on the same entity; the relationships, provenance and rules attached to it do not move. When a statutory definition changes, the mapping from the old category to the new one can be expressed once and applied across the whole inventory. When an auditor asks how an answer was produced, the evidence chain is the same whatever the technology is called that week.

/ 05What to watch

  • 28 November. The proposed legislative definition of Super Intelligence, and whether it departs from the current statutory definition of AI.
  • Agency guidance. How OMB and agency CAIOs implement the terminology in inventories, use-case reports and solicitations.
  • The accord in practice. Who the external auditors are, what they audit against, and whether any results are published.
  • Congress. Any new definition needs legislation to take full effect.
How it unfolded
SEP 29, 2026Six firms sign safety accord[2]SEP 29, 2026Executive Order 14434 signed[1]OCT 1, 2026This postNOV 28, 2026Legislative definition due[1]UPCOMINGSEP 29, 2026Six firms sign safety accord [2]SEP 29, 2026Executive Order 14434 signed [1]OCT 1, 2026This postNOV 28, 2026 · UPCOMINGLegislative definition due [1]
Aktiver’s take — Government

Renaming AI changes nothing an auditor can check. The 28 November definition might, because it could move systems in or out of every inventory, solicitation and reporting rule that relies on the statutory term.

Agencies that hold their use-case inventories as a graph can absorb that change in one edit. Each system is described by what it does, which data it uses and which decisions it informs; the statutory category is a classification attached to it, citing 15 U.S.C. § 9401(3) today and whatever replaces it later. A chief AI officer can then ask which use cases change status under the proposed definition, and which contracts reference them, before Congress acts. The same structure fits the accord’s four layers: internal controls, oversight findings, external audit reports and committee reviews linked to the model versions they examined, so a statement that a lab checked something points to the document showing it did.

How we’d model it

Use caseStatutory AI category15 U.S.C. 9401(3)Agency decisionAudit reportModel versionclassifiedAsdefinedByinformsexaminesUse caseStatutory AIcategory15 U.S.C.9401(3)AgencydecisionAudit reportModel versionclassifiedAsdefinedByinformsexamines
  • Use case —classifiedAs→ Statutory AI category
  • Statutory AI category —definedBy→ 15 U.S.C. 9401(3)
  • Use case —informs→ Agency decision
  • Audit report —examines→ Model version

See where this applies →

Sources — read the originals
  1. Inaugurating the Era of Super Intelligence — The White House whitehouse.gov ↗
  2. White House unveils 'super intelligence' executive order and industry accord — Nextgov/FCW nextgov.com ↗
Go deeper

See how Aktiver turns raw data into knowledge graphs your AI can reason over — deterministic, traceable, behind your firewall.

← All insights Subscribe via RSS → Book a demo →
Keep reading