TL;DR: On September 29, 2026, President Trump signed an order titled "Inaugurating the Era of Super Intelligence." It tells federal executive agencies to say "Super Intelligence" and "SI" instead of "artificial intelligence" and "AI" in official communications. For now it defines the new term by pointing to the existing federal definition of AI, so no statute, state law or private contract changed. The text we read sets no duty for private companies. The same day, six AI companies signed a voluntary Joint Commitment on Frontier Responsibilities with four layers of controls and no reported enforcement. The practical risk for a small team is two vocabularies in one compliance file. Below: a crosswalk of where each term applies, a contract alias clause, four vendor questions, and the November 28 date to watch.
Most of the coverage of this order is about the joke. A rename of AI is easy to mock, and several outlets did. If you run compliance for a small team, the useful question is narrower: does anything in your file need to change? Our answer after reading the order text on whitehouse.gov and four independent reports is that nothing legally changed for you, but your documents may now disagree with the documents you receive from federal agencies. That gap is where mistakes happen.
What the order says
The White House published the order, "Inaugurating the Era of Super Intelligence," as a presidential action dated September 29, 2026. Its operative parts are short.
- Who it binds: executive departments and agencies. They are to use "Super Intelligence" and "SI" in official correspondence, public communications, websites, reports and policy documents, to the extent the law allows.
- What it leaves alone: it does not require rewriting previously issued regulations, Presidential actions, contracts, grants or other historical documents.
- How it defines the term: for now, Super Intelligence means the technologies already covered by the federal statutory definition of artificial intelligence at 15 U.S.C. 9401(3).
- The 60-day task: the assistant to the president for science and technology must submit proposed legislative language for a new federal definition of Super Intelligence. That submission is to assess whether the definition should modify, expand on or supersede the current legal definition of AI, and may recommend further executive action.
- General provisions: the order creates no enforceable rights, and implementation depends on applicable law and appropriations.
Sixty days from September 29 lands on November 28, 2026. That is the date to put in your calendar, and it is the one place where this story could stop being cosmetic.
The Guardian reports that Trump made a similar declaration at the United Nations the week before. We did not verify that separately, and nothing in our analysis depends on it.
What it does not change
We read the order for any provision aimed at private companies and found none. That matches the International Association of Privacy Professionals summary, which notes the order does not rewrite existing laws.
So, concretely:
- The definition of an AI system in 15 U.S.C. 9401(3) is unchanged.
- State laws that regulate AI, including Texas TRAIGA (in effect since January 1, 2026) and Colorado's replacement law SB 26-189 (effective January 1, 2027), still use their own definitions. A federal executive order about agency vocabulary does not edit a state statute. This is our reading of how the order is scoped, not a quote from it.
- The EU AI Act is EU law and is untouched.
- Your existing vendor contracts, data processing agreements and policies mean what they said on September 28.
If you were looking for a reason to rewrite your AI policy this week, this is not it. Our federal preemption explainer makes the same point about the March 2026 White House framework: federal signals are not federal law, and state obligations stay in force until a court or Congress says otherwise.
The SI and AI crosswalk
The IAPP's observation is the one worth acting on: organizations will meet "SI" in federal materials while statutes, state laws, contracts, internal policies, vendor questionnaires and governance frameworks keep saying "AI." Here is where each term should show up, and how sure we are.
| Where you meet it | Vocabulary now | Basis |
|---|---|---|
| Federal agency websites, reports, guidance, correspondence | SI expected | Order text: applies to agency official communications |
| 15 U.S.C. 9401(3) and other federal statutes | AI, unchanged | Order points to the statute; only Congress can change it |
| Federal regulations, contracts and grants issued before the order | Unchanged | Order text: no rewrite required |
| State AI laws (Texas, Colorado, California, others) | AI, unchanged | Our reading: order is limited to the executive branch |
| EU AI Act and EU guidance | AI, unchanged | Not US federal law |
| Your vendor contracts, DPAs, security questionnaires | Whatever you wrote | No change in meaning from this order |
| Your internal AI policy and inventory | Your choice | No obligation to adopt SI |
The failure mode is not a regulator fining you for saying AI. It is a definitions argument. Suppose a vendor tells you in a renewal call that its new product is "SI, not AI, so your AI addendum does not apply." Or a federal customer sends a questionnaire that asks about "SI systems" and your answers refer only to "AI systems." Neither is a legal change. Both are avoidable with one sentence in your documents.
The Joint Commitment, in plain terms
The same day, the White House announced the Joint Commitment on Frontier Responsibilities. Trump signed it along with leaders of six companies, as reported by Nextgov and Al Jazeera:
- Anthropic (Dario Amodei)
- OpenAI (Greg Brockman, President)
- Google (Sundar Pichai)
- Meta (Mark Zuckerberg)
- xAI (Elon Musk)
- Nvidia (Jensen Huang)
Reports describe four layers of controls and audits:
- Internal monitoring of model capabilities and alignment during training and deployment.
- An internal team that checks the first layer is working and that issues are fixed.
- Independent external auditors who can assess the safety controls.
- An independent board or committee that reviews reports about those controls.
The companies also agreed to meet regularly to set standards and best practices. The commitment is voluntary. The reports we read describe no enforcement mechanism, no penalties and no deadlines. The Guardian notes that none of the layers involves government regulators, that companies can choose their own evaluators and oversight boards, and that there is no commitment to publish what the evaluations find. Al Jazeera adds that the document says it may make sense to codify measures into law later.
Reactions were split along predictable lines. Per Nextgov, Vice President JD Vance said the answer to AI risk is for companies to take it seriously rather than come to government for a regulatory regime. Al Jazeera's report quotes UNSW's Toby Walsh questioning an industry that checks its own work, and David Krueger of the University of Montreal calling the document too vague to matter much.
We did not read the commitment text itself, only reports of it. If you need to rely on any specific wording, get the primary document first.
Our position: treat this as a signal, not a control. A pledge with self-selected auditors and no reporting duty gives you nothing to point to in your own risk file. The reliable thing is still a contract term. Our AI safety body piece and the Amodei binding-regulation analysis cover why the argument over binding versus voluntary rules matters for vendor risk.
Four vendor questions that map to the four layers
If you buy from a frontier model provider, the commitment gives you a ready-made set of questions. Ask them whether or not the vendor signed, and write down the answers.
- Monitoring: how do you monitor model behavior in training and deployment, and what is the most recent category of issue it caught?
- Internal verification: which team verifies that monitoring works, and who does it report to?
- External audit: which independent auditor reviews your safety controls, what was the scope, when was the last review, and can we see findings under NDA?
- Oversight: who sits on the independent board that reviews audit reports, and who appoints them?
Then add the question that matters most for your risk file: does anything in our contract change because of this commitment? If the answer is no, the pledge is marketing for your purposes. For the full diligence set, use the AI vendor due diligence checklist and the agentic AI contract clauses.
Copy-paste language
Add this alias line to your AI policy or glossary section:
Terminology. In this policy, "AI system" and "artificial intelligence" include any
system that a government agency, vendor or marketing material describes as "Super
Intelligence" or "SI". Where a statute or regulation defines these terms, that
definition controls.
Add this to the definitions section of vendor agreements at renewal:
"AI System" means any machine-based system within the meaning given in Section [X]
of this Agreement. A system is an AI System regardless of whether Provider, any
government body or any marketing material labels it "artificial intelligence", "AI",
"Super Intelligence", "SI" or any successor term. Provider may not rely on a change
in labeling to avoid obligations that apply to AI Systems under this Agreement.
Add this to your vendor questionnaire:
Does Provider describe any product that it supplies to us as "Super Intelligence" or
"SI"? If yes, confirm in writing that each such product is an AI System for the
purposes of our agreement, our AI policy and any applicable law.
You can browse the current definition work in our glossary entry for Super Intelligence and the regulation entry for the order.
What to do this week
- Do not rename anything. Keep citing statutes and state laws by the terms they use.
- Add the alias line to your AI policy and your vendor contract template.
- Update your questionnaire so federal customers' "SI" language and your "AI" language point to the same systems.
- Calendar November 28, 2026 for the proposed statutory definition, and read it when it appears.
- Ask your frontier-model vendors the four questions and file the answers with your vendor record.
For policy language that already covers AI use by staff, our acceptable use policy template is the starting point.
What we do not know yet
- What definition will be proposed. The 60-day task asks whether the new definition should modify or supersede the current one. A broader or narrower definition could change which systems statutes cover. Nobody outside the White House has seen the language.
- How agencies will handle the transition. The order exempts historical documents but does not say how fast current ones are rewritten.
- The commitment's full text and follow-up. We saw reports, not the document. We have no information on whether the reported meetings will happen or be public.
- Whether other companies will join. Reports name six signers. If your vendor is not one of them, the pledge says nothing about it.
- Small differences between outlets. For example, reports list OpenAI's signer as its President, Greg Brockman. We used the names that two or more outlets agreed on.
We will update this page when the November 28 submission appears.
Related Reading
- Federal AI preemption 2026: which state AI laws still apply
- AI safety body: FAA, FINRA or IAEA model, three vendor risk questions
- Superintelligence ban bills: four enterprise governance questions
- Dario Amodei AI regulation proposals: binding rules and vendor risk
- OpenAI's new deal policy: three governance steps for small teams
- Agentic AI vendor contracts: the eight clauses you need
- AI vendor due diligence checklist 2026
- Newsom AI kill switch order N-9-26: four vendor contract clauses
- Who enforces when your AI vendor fails: DOJ and state AGs
- AI acceptable use policy template for small teams
