# White House Accord on Super Intelligence, Explained

- URL: https://superintelligence-news.com/articles/white-house-accord-super-intelligence-explained
- Publisher: Superintelligence News
- Section: Policy & Government
- Type: explainer
- Author: Umair
- Published: 2026-10-19
- In this story: [Donald Trump](https://superintelligence-news.com/people/donald-trump), [OpenAI](https://superintelligence-news.com/companies/openai), [Anthropic](https://superintelligence-news.com/companies/anthropic)

**In plain English:** The pledge asks people inside a company, an outside reviewer and the board to check that the company's AI safeguards work.

> The White House AI accord is a voluntary pledge, not a new law. Six company executives and President Donald Trump signed it on September 29, 2026. It calls for checks within companies, outside review and board oversight. Existing laws governing misleading business claims remain separate from the pledge.

## The brief

- **What happened:** Six firms signed at a White House gathering, but lunch guests from Microsoft, Amazon and Palantir did not.
- **Why it matters:** ABC News found it unclear whether the pact adds to outside-review promises OpenAI and Anthropic had already made.
- **What happens next:** The signers plan regular standards meetings. Trump also floated a possible 10-person oversight committee.

## Key facts

- Forbes reported on September 30, 2026, that the White House AI accord contains 308 words.
- House Speaker Mike Johnson announced no congressional action to enforce the accord's principles after the September 29, 2026, signing, ABC News reported.
- The September 29, 2026, accord says its signers remain committed to the controls whether or not the law requires them, according to the American Presidency Project's text.

President Donald Trump and executives from six tech companies signed the White House AI accord on September 29, 2026. It calls for four layers of checks on advanced AI systems. The pledge is voluntary, with no penalties or enforcement process of its own.

## What is the White House AI accord?

The White House AI accord is a pledge by six companies to check how their advanced AI systems behave. Its subtitle is "Joint Commitment on Frontier Responsibilities". Frontier AI means the most advanced AI systems. The text puts responsibility for safe development on each company, not a new regulator.

The title uses "Super Intelligence" (SI), the US government's new name for AI. The Center for AI Safety says the phrase refers broadly to AI here. The name does not mean AI has become superintelligent. Our [explainer on the government's AI label](/articles/trump-rename-ai-super-intelligence) covers the naming issue.

CBS News based its account on the copy Trump posted to Truth Social. The American Presidency Project carries the accord's text and signature list.

## Which companies signed the White House AI accord?

Executives signed for Google, Anthropic, Meta, OpenAI, Nvidia and xAI on September 29, 2026, alongside President Donald Trump. Other lunch guests came from Microsoft, Amazon and Palantir. Those guests did not sign the document, according to General Legal's guide to the accord.

| Company | Executive who signed |
| --- | --- |
| Google | Sundar Pichai, chief executive |
| Anthropic | Dario Amodei, chief executive |
| Meta | Mark Zuckerberg, chief executive |
| OpenAI | Greg Brockman, president |
| Nvidia | Jensen Huang, chief executive |
| xAI | Elon Musk |

General Legal names the guests who did not sign as Microsoft's Satya Nadella, Amazon founder Jeff Bezos and Palantir's Alex Karp. CNBC identifies Brockman as OpenAI's signer and president.

The Next Web, citing Semafor and Politico, reported that Zuckerberg came up with the accord and circulated a draft before lunch. That account attributes the idea to Zuckerberg; it is not a statement from Meta.

## What are the accord's four layers of controls?

The accord asks companies to build four levels of review. They span checks within AI systems, an in-house team, an outside auditor and an independent board committee. These layers cover finding problems and making sure they get fixed. The firms also promise to discuss shared safety practices.

1. Model checks. Track what AI systems can do and whether they behave as intended while being trained and used. The text names cybersecurity, biological and chemical threats, and unwanted access to computer systems.
2. In-house review. Give a team the power to check whether controls and warning systems work, and ensure problems are fixed.
3. Outside review. Bring in an independent auditor or evaluator to assess those same controls and warning systems.
4. Board review. Set up an independent board committee to oversee the work and receive findings from the teams and reviewers. The committee must also make sure identified problems are fixed.

The pact's wording recommends these steps. It does not set a legal requirement to carry them out.

ABC News reported that OpenAI and Anthropic had already promised to use outside evaluators. It was unclear whether the accord added to those promises. The Center for AI Safety reached a similar view in its newsletter, judging the commitments to add little to existing pledges.

Anthropic's [METR access program](/articles/anthropic-metr-employee-like-access) gives one example of an outside review arrangement. Anthropic said six METR employees received company hardware, office space and access to internal tools. It compared their access with that of its own risk assessment teams.

The company also said it cannot block METR from publishing its findings, though it can require cuts under set rules. Anthropic decides which information falls within those rules. METR retains a right to comment on the access and information it received.

## Is the White House AI accord legally binding?

No. The accord itself is not a law and creates no penalty or enforcement process, according to General Legal's review. It also stops short of naming a regulator or deadline. Its wording urges companies to take the steps rather than requiring them by law. House Speaker Mike Johnson called the industry commitments voluntary.

At the White House on September 29, 2026, Trump told reporters: "I think it's morally binding." CBS News reported his answer to a question about the agreement's force.

Existing law is a separate issue. General Legal notes that the FTC Act's ban on deceptive practices and state consumer-protection laws apply to most companies. Those rules concern what firms say and do, not a new enforcement power created by this pact.

General Legal also cites California's SB 53, in effect since January 1, 2026. The guide describes a ban on materially false or misleading claims by frontier developers. That ban covers catastrophic AI risk and how developers manage it. The guide notes an exception for reasonable statements made in good faith.

Crowell & Moring warned that the planned company meetings could raise antitrust concerns, or questions about unlawful coordination between rivals. That is a warning about possible legal issues, not a finding that the accord breaks the law.

## What happens next for the White House AI accord?

The next step written into the accord is regular company meetings to develop shared safety practices. The text allows for later laws or regulations but does not require them. Trump also said executives discussed a 10-person oversight committee. His remarks described a possible body, not an established one.

Some signers had sought binding rules before the signing. In June 2026, Amodei called for tougher, binding AI regulation, Al Jazeera reported. In September 2026, before the pact, OpenAI chief global affairs officer Chris Lehane urged Congress to set mandatory national standards. He said industry standards should sit alongside federal safeguards and democratic oversight, not displace them.

Democratic US Senator Mark Warner of Virginia also sought laws after the lunch, CBS News reported. He called for required tests and reviews of the most advanced AI systems, plus duties to report incidents. Those were requests for legislation, not rules enacted by the accord.

The accord sets no date for its first standards meeting. The White House has set up a [Super Intelligence Force](/articles/super-intelligence-force) of senior officials, which at its first meeting on October 15, 2026 created a working group to see that the accord is put into practice.

## Frequently asked questions

### Did Microsoft or Amazon sign the White House AI accord?

Neither company appears in the signature block. General Legal identifies Microsoft's Satya Nadella and Amazon founder Jeff Bezos as lunch guests who did not sign. Their attendance did not make either company a signer.

### Does the accord require public audit reports?

Al Jazeera reported that publishing audit findings is not one of the pact's requirements. An independent review and a report that anyone can read are separate steps.

### Does the accord set a technical testing standard?

General Legal found no standard in the document itself. The pact assigns oversight roles but does not specify a test method, a passing score or a chosen outside auditor.

### Does the title mean AI has become superintelligent?

No. The Center for AI Safety explains the wording as Trump's broad term for AI. The accord concerns checks on the technology; its name is not evidence of a new level of ability.

## Sources

- [White House Accord on Super Intelligence — full text](https://www.presidency.ucsb.edu/documents/white-house-accord-super-intelligence)
- [Understanding the White House Accord on Super Intelligence: A Guide for AI Companies](https://general.legal/blog/white-house-accord-on-super-intelligence-a-guide-for-ai-companies)
- [Trump and major AI executives sign morally binding voluntary controls](https://www.cbsnews.com/news/trump-ai-constitution-tech-execs-openai-anthropic-voluntary-controls/)
- [Trump says AI leaders signed a constitution to police themselves](https://abcnews.com/Politics/top-ai-leaders-meet-trump-white-house-amid/story?id=136832988)
- [How does Trump's White House AI accord work?](https://www.aljazeera.com/economy/2026/9/30/how-does-trumps-white-house-ai-accord-work)
- [Trump tries to rename AI super intelligence as polls show him sinking on key issue](https://www.cnbc.com/2026/09/29/trump-ai-super-intelligence.html)
- [An update on independent evaluators](https://anthropic.com/news/an-update-on-independent-evaluators)
- [AISN #83: Six AI Companies Sign White House Accord](https://newsletter.safe.ai/p/aisn-83-six-ai-companies-sign-white)
- [White House AI Responsibility Accord & Super Intelligence Task Force: What Businesses Need to Know](https://www.crowell.com/en/insights/client-alerts/white-house-announces-ai-responsibility-accord-executive-orders-and-super-intelligence-task-force)
- [Zuckerberg drafted Trump's AI accord, but Microsoft and Amazon didn't sign](https://thenextweb.com/news/trump-ai-accord-zuckerberg-microsoft-amazon-didnt-sign)
- [Here's the White House's 308-Word AI Accord in Full](https://www.forbes.com/sites/saradorn/2026/09/30/white-house-releases-accord-between-billionaire-ai-execs-heres-what-it-says/)