The White House Super Intelligence Accord: Who Signed and What It Says

Updated 1 Oct 2026 · Reviewed by the Zekai editorial team

The accord was signed at the White House on the same day as the executive order that renamed AI to SI in federal language. Because both carry the words “super intelligence”, many people assume they are one document. They are not. The order changes words used by government agencies. The accord is a set of promises by private companies about how they check their most powerful models.

What is the White House Accord on Super Intelligence?

The full title is the White House Accord on Super Intelligence: Joint Commitment on Frontier Responsibilities. It is a short statement of principles about “frontier models”, the largest and most capable AI systems. The text uses “should” rather than “must”, and President Trump described it as “morally binding” rather than legally binding.

Who signed the super intelligence accord

Swipe the table sideways →

Who did not sign

Microsoft’s Satya Nadella, Amazon’s Jeff Bezos, AMD’s Lisa Su and Palantir’s Alex Karp attended the White House event but did not sign. Apple was not among the signers either. None of them has explained publicly why, so treat any reason you read as a guess.

What the accord commits companies to

The accord describes four “layers” of oversight for frontier models:

1. Internal technical controls that monitor what a model can do during training and after release, with a focus on cybersecurity, biosecurity and chemical-weapon risks, and on stopping models from “hacking or accessing technical systems in unintended ways”.
2. An internal oversight team whose job is to check that those controls actually work.
3. An independent external auditor that reviews whether the safeguards do what the company says.
4. An independent board committee that reads the internal and external findings and makes sure problems get fixed.

What the accord does not do

• No penalties. Nothing happens to a company that ignores it.
• No definition of a frontier model. Each company decides which of its models count.
• No named auditors, standards or deadlines.
• No public reporting. Outside auditors report to the companies’ own boards, not to regulators or the public.
• No new licensing or pre-approval. Legal analysts note it creates no new mandatory obligations.

Supporters see it as a first shared baseline that the 60-day legal review under the executive order could later turn into real rules. Critics, including Senator Mark Warner, say it does little about the actual risks of more powerful systems.

The “Unites States” typo

If you saw jokes about “Unites States” this week, they come from the signed accord itself: the signature block misspelled “United States”. The error was spotted by reporters and spread quickly online. It has no legal effect, because the accord is voluntary anyway.

Accord vs executive order: the difference

Swipe the table sideways →

What the accord means if you use these tools

Day to day, nothing changes. ChatGPT, Claude, Gemini and the other products work exactly as before, at the same prices. The accord is about how companies test and supervise their next large models before and after release. If the promises are kept, the main effect for users would be stricter limits on dangerous requests (cyberattacks, weapons) and slower or more cautious releases of the biggest models. If you need a tool from a company outside the accord, our SI tools list and the full directory cover alternatives in every category.

▶ Watch the Claude video review (0:37)

▶ Watch the ChatGPT video review (0:31)

Frequently asked questions