From AI to SI: what the US super intelligence order means for UK businesses

On 29 September 2026, President Trump signed an executive order instructing US federal agencies to use the term "super intelligence", or "SI", instead of "artificial intelligence" in official correspondence, websites, reports and other non-statutory documents, as the order itself sets out. The same day, six AI companies signed a voluntary safety accord with the US government. For a UK small business, neither changes a single law, contract or data protection duty you already work under. What they do change is worth ten minutes, so you know what to ignore and what is worth acting on.

What does the super intelligence executive order actually do?

Executive Order 14434 tells the US executive branch to swap "Artificial Intelligence" and "AI" for "Super Intelligence" and "SI" in official correspondence, public communications, websites, reports and policy documents, "to the maximum extent permitted by law", according to the order itself. It does not touch previously issued regulations, contracts or grants. For now the order defines SI as covering exactly the technologies already captured by the existing US statutory definition of AI, so nothing about what counts as the technology has changed. What has changed is that the Assistant to the President for Science and Technology now has 60 days to propose legislative language for a formal federal definition of SI, including an assessment of whether it should "modify, expand upon, or otherwise supersede" the legal definition of AI, as Nextgov reported. That is a genuinely open question in the US, not a settled fact.

Does this change the law in the UK?

No. The order instructs the US executive branch about its own wording. It has no force in the UK and creates no obligation for any UK business, whatever tool you use. In practice, expect US vendors to start using "super intelligence" or "SI" in their own marketing, because that is now the preferred term inside the administration they are courting. Treat it as you would any rebrand: a new name for the same technology, not evidence that a tool is safer, smarter or more compliant than it was last month.

What is in the AI safety accord, and does it bind anyone?

On the same day, the White House announced a voluntary accord on frontier model safety, signed by Google's Sundar Pichai, Anthropic's Dario Amodei, Meta's Mark Zuckerberg, OpenAI's Greg Brockman, xAI's Elon Musk and NVIDIA's Jensen Huang alongside Trump, with Infosecurity Magazine carrying the fullest breakdown. It sets out four layers of controls: monitoring during training and deployment, an internal team overseeing those controls, an independent external auditor, and a board level committee receiving reports from both. Vice President JD Vance pushed back on the idea of a new FDA or FAA style regulator for frontier models, favouring companies managing the risk themselves over a new regulatory regime, as Nextgov reported. The accord carries no legal penalty for breaking it. Trump called it morally binding and said AI firms should police themselves. Days later he announced a new Super Intelligence Force with 120 days to report on AI risks while working to prevent "overregulation", according to TechCrunch. The pattern is consistent: self governance over statutory rules, for now.

Why did the same companies just decline to guarantee their own safeguards?

Two events either side of the accord are worth knowing. The same day, OpenAI cancelled the release of a new model, GPT-6.1 Astra, after internal testing found it did not meet the company's own standard for staying within scope and acting on instructions as intended, according to Al Jazeera. Then in early October, representatives from OpenAI, Anthropic, Meta and Google appeared before a New York City Council hearing and were asked directly whether they could guarantee their AI agents would always follow safety guardrails. None would. OpenAI's Morgan Dwyer said it was "not possible... to commit or guarantee that any technology is without risk." Anthropic's Logan Graham called the underlying science "fundamentally hard and unsettled." Google's Alice Friend said "to promise perfection would not be possible with any product on the market," as Fox News reported. That is the vendors being candid, under direct questioning, about the limits of their own testing, a useful data point for anyone buying their tools.

What should a UK small business actually do about this?

Nothing here requires you to change a tool, a contract or a process this week. What it should change is how much weight you put on a vendor's own marketing. The checks on how AI is used inside your business were always yours to build, and this month's news just makes that harder to ignore: the biggest labs in the world are policing themselves voluntarily, and will not fully guarantee the result even then.

What to do this month

• Write down, even briefly, which AI tools are approved for use in your business and what they are not approved for. A one page AI policy beats no policy at all.
• Know which tool touches which type of data, especially customer or financial data, before you worry about which model powers it. Our guide on whether AI is safe for your business data covers the real questions to ask.
• Keep a person signing off anything customer facing or anything that moves money, regardless of how capable a tool claims to be.
• Check which tier you are actually on. Free, Business and Enterprise plans from the same vendor can have very different data handling terms; see our comparison of ChatGPT and Claude's free, business and enterprise tiers.
• If you want a second pair of eyes on where AI sits in your business, our AI Operations Day and compliance work both start there, and Claude for Business is worth a read if Anthropic's tools are in your stack.

None of this depends on what anyone decides to call the technology. Call it AI, call it SI, the job of checking it inside your own business is still yours.