Skip to content

Published · 11 min read

Understanding the White House Accord on Super Intelligence: A Guide for AI Companies

The White House Accord on Super Intelligence is a voluntary pledge, not a law. Even so, customers, investors, and counterparties may ask AI developers, and the companies that use their models, whether they follow its four layers of controls.

Victor MeerVictor MeerSenior Commercial Counsel

Accord and Executive Order

The White House Accord on Super Intelligence is a voluntary pledge of about 300 words, signed on September 29, 2026 by President Trump and six technology leaders. It asks companies building frontier-AI models to adopt four layers of safety controls and audits.

The six company signers were Sundar Pichai (Google), Dario Amodei (Anthropic), Mark Zuckerberg (Meta), Greg Brockman (OpenAI), Jensen Huang (Nvidia), and Elon Musk (xAI). Other executives attended the White House lunch but did not sign, including Microsoft's Satya Nadella, Amazon founder Jeff Bezos, and Palantir's Alex Karp.

The same day, the President signed a separate executive order, Executive Order 14434, "Inaugurating The Era Of Super Intelligence," which directs federal agencies to say "Super Intelligence" and "SI" instead of "artificial intelligence" and "AI" in official correspondence, public communications, websites, reports, policy documents, and other non-statutory documents. SI has the meaning that 15 U.S.C. 9401(3) gives "artificial intelligence," and the order does not require changes to previously issued regulations, contracts, or grants. You do not need to rewrite your contract definitions now.

The order does start a process that could change the definition of Super Intelligence. It gives the Assistant to the President for Science and Technology 60 days, until about November 28, 2026, to submit proposed legislative language to the President for a federal SI definition, along with an assessment of whether the new definition should modify, expand, or supersede the existing statutory definition of artificial intelligence.

Full Text of the Accord

The full text of the accord, as signed on September 29, 2026, reads:

White House Accord on Super Intelligence

Joint Commitment on Frontier Responsibilities

In order to build a positive future for the American people and the world, we believe every company is responsible for developing its own technology safely and in a way that builds trust with customers and the public.

This starts with every company that is training and deploying frontier models having robust internal processes and controls to ensure that their technology behaves as intended and that any issues are promptly identified and resolved. Therefore, in addition to any other precautions, we believe each company should implement the following four layers of controls and audits:

(1) Implement robust internal controls to monitor the capabilities and alignment of its models during training and deployment around areas like cybersecurity, biosecurity, and chemical threats, and to ensure that its models do not hack or access technical systems in unintended ways.

(2) Empower an internal team to ensure all of the controls, monitoring, and detection are operating as intended, and that any issues are remediated.

(3) Partner with an independent external auditor or evaluator to carry out independent assessments of whether the controls, monitoring, and detection are operating as intended.

(4) Designate an independent committee of the board of directors to oversee and receive reports from the teams operating the controls and the internal and external auditors and evaluators, as well as to ensure any issues identified are remediated.

Together, these steps will give each company, its customers, and the public confidence that the technology is operating as intended.

The participating companies will meet regularly to establish standards and best practices to improve the safety of their systems.

Over time, it may make sense to codify these steps into laws or regulations. Regardless of whether this is required of companies, we believe that implementing these controls and audits is critical to ensuring a safe future for everyone, and each of our companies are committed to doing this.

DONALD J. TRUMP, President of the Unites [sic] States · SUNDAR PICHAI, Google · DARIO AMODEI, Anthropic · MARK ZUCKERBERG, Meta · GREG BROCKMAN, OpenAI · ELON MUSK, XAI · JENSEN HUANG, Nvidia

The accord is not legally binding. Though the text says that codifying the steps "may make sense" over time, the current accord only suggests that each company "should" implement the accord's four layers. Its signers claim that their companies "are committed to doing this." President Trump called the accord "morally binding," and House Speaker Mike Johnson described the commitments as voluntary.

The accord names no regulator, penalty, deadline, standard, or enforcement mechanism. However, regulators and customers can compare what a company says about its controls with its later actions.

In California, SB 53, in effect since January 1, 2026, bars any frontier developer (a developer that has trained, or begun training, a model using more than 10²⁶ integer or floating-point operations) from making a materially false or misleading statement about catastrophic risk from its frontier models or its management of that risk, with an exception for statements made in good faith that were reasonable under the circumstances (Bus. & Prof. Code section 22757.12(e)). The statute's separate ban on false statements about compliance with a developer's own published safety framework, and its civil penalties of up to $1 million per violation, apply only to large frontier developers (Bus. & Prof. Code sections 22757.12(e)(1)(B) and 22757.15), meaning those with more than $500 million in annual gross revenue. A frontier developer that claims to follow accord-style controls should thus be able to prove it. Meanwhile, the FTC Act's prohibition on deceptive practices (15 U.S.C. 45) and state consumer-protection laws apply to most companies.

Four Layers of Controls

The accord asks each company that trains and deploys frontier models to put four layers of controls and audits in place. In the accord's order:

  1. Internal controls. Monitor model capabilities and alignment during training and deployment in areas such as cybersecurity, biosecurity, and chemical threats, and make sure models do not hack or access technical systems in unintended ways.
  2. An internal owner. Empower an internal team to confirm that the controls, monitoring, and detection work as intended, and to fix problems that turn up.
  3. Independent outside review. Partner with an external auditor or evaluator to assess whether those controls work as intended.
  4. Board oversight. Designate an independent committee of the board of directors to oversee and receive reports from the teams operating the controls and from the internal and external auditors and evaluators, and to make sure identified issues are remediated.

The signers also agreed to meet regularly to develop standards and best practices. The text does not define "frontier models," name an auditor or a standard, set a deadline, or attach a penalty.

Effects on Companies

By its terms, the accord's four layers apply to companies "training and deploying frontier models." Depending on your position in the AI supply chain, this will have different implications for your company. Signers have made a voluntary pledge while frontier developers that did not sign may hear from customers and investors asking whether they have internally committed to follow the same controls. Other companies building on third-party models are not covered, but they may now raise more questions to their model providers. Finally, companies selling to federal agencies should read the executive order governing the terminology used by agencies.

Because the accord does not define "frontier," a company training large models should never assume that it is excluded. SB 53's threshold of 10²⁶ operations is one reference point, and New York's RAISE Act, which takes effect January 1, 2027, uses the same threshold.

Effects on the Broader AI Industry

Despite creating no new legal mandate, the accord still has meaningful implications.

It signals a federal preference for self-regulation, but Congress and federal agencies remain active. The President told reporters he was seeing "tremendous self-policing," and Vice President JD Vance said developers should take responsibility for their products instead of asking the government to regulate them. The House FRONTIER Act (H.R. 9925) from Reps. Obernolte and Trahan, introduced in July, would require independent audits of large frontier developers and would preempt some state frontier-AI laws. Congress is now out of session until after the November election. In the Senate, talks among Senators Thune, Cruz, and Klobuchar on a catastrophic-risk bill have stalled, no bill text has been released, and Cruz has said any markup would have to wait for the post-election session. On September 29, Senator Cruz also blocked a Democratic attempt to pass a separate bill creating an AI Safety Board, the Artificial Intelligence Risk Management and Security Act of 2026, by unanimous consent. The FTC also confirmed on September 30 that it is investigating OpenAI, Anthropic, and other AI companies over possible risks to consumers. A senior FTC official told the New York Post the agency plans to issue formal demands for information and that the probe began before OpenAI agents hacked Hugging Face in July. On October 4, the President named Director of National Intelligence Jay Clayton to chair a new Super Intelligence Force, which will coordinate the federal government's engagement with AI companies and other groups. FTC Chairman Andrew Ferguson is one of its vice chairs, and the group has 120 days to report on AI risks and opportunities.

The accord's four layers could become an accepted rubric. Internal monitoring, a named owner, independent review, and board reporting are easy to turn into questionnaire items. Enterprise customers, investors, and acquirers may start asking vendors and portfolio companies how their controls compare.

Parties may define the accord's open terms in their contracts. The accord does not define "independent," "as intended," or "frontier." In a negotiation, the party with more leverage often decides what those words mean. If you buy AI, define them in your favor. If you sell AI, make sure you clarify definitions before agreeing to accord-style controls.

Many companies would benefit from reviewing their current position.

  1. Define your position in the AI supply chain. Determine whether you are a frontier developer, a company building on someone else's models, or a buyer of AI tools. This helps determine what controls are relevant to you.
  2. Compare your controls with the four layers. Check whether you monitor model behavior, have a named owner, get independent review, and report to the board. Even smaller companies should be aware of their gaps.
  3. Verify against what you publicly claim. Trust pages, security questionnaires, investor decks, and customer contracts should match what you do on AI oversight.
  4. Ask your model providers about their commitments and put them in the contract. You may want to cover incident notice, evidence of independent assessment, and definitions of "independent" and "frontier."
  5. Mark your calendar. Three dates matter: the proposed federal SI definition is due to the President around November 28, 2026; the Super Intelligence Force's report is due within 120 days, roughly early February 2027 if the clock started with its October 4 announcement; and New York's RAISE Act takes effect January 1, 2027. Also watch for Senate action on a frontier AI bill after the election.

General Legal's Technology Transactions practice can help you turn the accord into practical steps. We can map your controls against the accord's four layers, review AI oversight items on your website, questionnaires, and customer agreements, and update vendor and customer terms to match. If you sell to the federal government, our Defense Tech and Government Contracts practice can review how the SI terminology appears in agency solicitations and in your proposals. To have us review where your company stands, contact General Legal to schedule a consultation.

This post reflects a general legal framework and is not legal advice for any specific company's situation.

Sources

Key takeaways
  • The White House Accord on Super Intelligence is a voluntary 300-word pledge signed September 29, 2026 by President Trump and six tech leaders, not a legally binding regulation.
  • The accord asks frontier-model developers to implement four layers of controls: internal monitoring, a designated internal team, independent external audit, and board oversight.
  • A concurrent executive order directs federal agencies to use "Super Intelligence" and "SI" terminology instead of "AI," but does not change legal definitions or require contract rewrites.
  • While not enforceable on its own, the accord's commitments may become scrutinized under existing laws like California SB 53, which bans materially false statements about catastrophic AI risk management.
  • Companies should verify that public claims about AI safety controls match actual practices, and buyers of AI should ask providers about accord-style commitments in contracts.
  • Key upcoming dates include the proposed federal SI definition due around November 28, 2026, the Super Intelligence Force report in early February 2027, and New York's RAISE Act effective January 1, 2027.
TL;DR
Legal StatusThe accord is a voluntary pledge with no regulator, penalty, or enforcement mechanism, though signers say they are committed to its terms.
Four Control LayersInternal controls for model monitoring, an empowered internal team, independent external audit, and board-level oversight committee.
Who Is CoveredBy its terms, companies training and deploying frontier models; no threshold is defined, but California's SB 53 uses 10²⁶ operations as a reference.
Executive Order 14434Agencies must say "Super Intelligence" instead of "AI" in official documents; a proposed federal definition is due around November 28, 2026.
Existing Legal RiskCalifornia SB 53 and the FTC Act already bar false statements about AI safety; public claims about accord-style controls should match actual practices.
Practical ImplicationsCustomers, investors, and counterparties may ask whether companies follow the four layers, even if not legally required.
Industry DevelopmentsCongress is working on the FRONTIER Act requiring audits; the FTC is investigating OpenAI and Anthropic; a new Super Intelligence Force will coordinate federal engagement.
Recommended ActionsMap your controls to the four layers, verify public statements match reality, and define accord terms in vendor and customer contracts.
FAQs

Is the White House Accord on Super Intelligence a law?

No. It is a voluntary pledge in which the signers say their companies are committed to four layers of controls and audits, and that codifying them "may make sense" later.

Does my company have to follow the four layers?

Not under the accord, which has no enforcement mechanism of its own. However, customers, investors, and counterparties may ask about similar controls, and your public statements about them should be accurate.

Did "AI" legally become "Super Intelligence"?

No. The executive order changes how the executive branch writes and speaks, not the legal definition. A proposed federal definition is due about November 28, 2026.

What should my company do first to align with the accord?

Make sure what you say publicly about your AI controls matches what you do, then check what your model providers have promised you in writing.

Related reading