White House accord on super intelligence: who signed, what they promised, and the commitments log

The White House Accord on Super Intelligence is a one-page voluntary pledge that six companies signed on September 29, 2026, the same day the President signed the order renaming AI. This page is a living tracker: it records who signed, exactly what they promised, what the pledge leaves out, and a dated log we will extend as signers act. Every entry carries its source.

A caveat first. We could not open the accord document on whitehouse.gov, where the executive order and fact sheet are posted, and the fact sheet we read does not mention the accord. The wording below comes from outlets that reproduced or closely summarized it, and we say which. When the primary text is published at a stable address we will replace these with direct quotes.

What was signed

The document is titled "White House Accord on Super Intelligence, Joint Commitment on Frontier Responsibilities," according to the Washington Examiner's report; MLex gives the same pledge as "Joint Commitment on Frontier SI Responsibilities." It carries seven names, as reported by IBTimes and Luiza Jarovsky's newsletter:

• Donald J. Trump, President
• Sundar Pichai, CEO, Google
• Dario Amodei, CEO, Anthropic
• Mark Zuckerberg, CEO, Meta
• Greg Brockman, President, OpenAI
• Elon Musk, founder, xAI
• Jensen Huang, CEO, Nvidia

Microsoft and Amazon do not appear in any signatory list we read.

The four layers

According to IBTimes' reading of the text, each company commits to:

1. Internal controls. "Robust internal processes and controls" to monitor model capabilities and alignment during training and deployment, covering cybersecurity, biosecurity and chemical threats. 2. An internal oversight team. An empowered team that makes sure the controls and detection systems work and that problems are addressed. 3. An independent external auditor. Outside evaluators who assess whether the controls and monitoring work as intended. 4. An independent board committee. A board-level committee that receives reports from the control team and the auditors and ensures issues are fixed.

The accord also states that it "may make sense to codify these steps into laws or regulations" over time, while calling implementation critical regardless of legal requirements, per IBTimes.

What it is, and is not

Trump described the pledge as "morally binding," as relayed by the outlets above. That is a description, not a legal status. Freshfields states the accord "does not create legally enforceable obligations or an enforcement mechanism," and adds that "the accord does not expressly require signatories to adopt the Administration's rebranding." The reading from IBTimes and from privacy lawyer Luiza Jarovsky's analysis goes further on the gaps:

• No penalties and no government enforcement role. Nothing happens to a company that falls short.
• Undefined terms. "Robust" and "operating as intended" are not defined, so there is no yardstick for compliance.
• Company-picked auditors. Jarovsky writes that the "independent" auditor is chosen and paid for by the company, and that the board committee sits inside the company.
• No disclosure. Nothing requires publishing auditor findings or board reports.

Jarovsky argues the accord amounts to endorsing self-regulation, and she points to the OpenAI and Hugging Face security breach as the kind of incident it is unlikely to prevent. That is her analysis, and we note it as such.

The commitments log

Dated entries only. "Not yet reported" means nothing verified yet.

• September 29, 2026: Accord announced and signed by six companies and the President (IBTimes, Luiza Jarovsky).
• September 29, 2026: Executive order renames AI "Super Intelligence" in federal documents; the accord does not require signers to follow (Freshfields).
• About November 28, 2026: Science adviser's proposed statutory definition due under the order (executive order, via Freshfields).
• Not yet reported: names of external auditors per signer.
• Not yet reported: board committee charters or first reports.
• Not yet reported: any enforcement or codification step.

For the order's mechanics, see our explainer on the executive order. For what the central word means, see what is superintelligence.

Where it sits in the policy story

The fact sheet for the order lists the administration's own framing: a $5 billion Genesis Mission for applying these systems to science, medicine, energy and manufacturing, the July 2025 "America's AI Action Plan" with more than 90 federal actions, and a March 2026 "National Super Intelligence Policy Framework." Those are the administration's descriptions and we have not checked the figures. The accord adds an industry-side layer to that stack without adding a government one. The distinctive feature is its structure: four named layers, with an outside auditor and a board committee. That is more specific than a general promise to act responsibly. The weakness, as the critics above note, is that the company controls the inputs and the publication.

What would change our verdict

We would upgrade this from a pledge to a mechanism if any of these appear: a signer naming its auditor and the scope of the audit; a signer publishing a board committee charter; a published summary of an audit finding; a statute or rule that makes any layer mandatory. We would downgrade it if a signer reports an incident the controls should have caught and gives no account of it.

Our take

The accord is real, signed, and more specific than most self-regulation, but it asks nothing that outsiders can check. Treat the next ninety days as the test: do the six companies name auditors and publish anything? We will add each dated event to the log above, and we will re-send this page, with a correction note, when the primary text is available to quote.