President Donald Trump signed an executive order on September 29 requiring federal agencies to replace "Artificial Intelligence" with "Super Intelligence" across their official documents, hours after six frontier labs signed a voluntary safety accord at the White House that names no audit schedule and imposes no penalty for ignoring it.
Key takeaways
- The order tells agencies to use "Super Intelligence" and "SI" in correspondence, websites, reports and policy documents and to stop acknowledging "AI" — but Section 3 keeps the existing statutory definition at 15 U.S.C. 9401(3) in force until Congress acts.
- Trump signed the White House Accord on Super Intelligence alongside Sundar Pichai, Dario Amodei, Mark Zuckerberg, Greg Brockman, Elon Musk and Jensen Huang, committing Meta, Google, OpenAI, Anthropic, Nvidia and xAI to four layers of controls and audits.
- The one-page document never defines what makes internal controls "robust" and never says how often external auditors or evaluators should inspect a signatory's systems.
What the executive order actually changes
The order's operative section is about vocabulary. Agencies must swap in "Super Intelligence" and "SI" for official correspondence, public communications, websites, reports, policy documents and other non-statutory material. It explicitly does not require rewriting previously issued regulations, presidential actions, contracts, grants or historical documents.
The scope of the new term is, for now, identical to the old one. Section 3 borrows the definition of artificial intelligence already written into federal law and says it governs until superseded by Congress or a later presidential action. The Assistant to the President for Science and Technology has 60 days to propose legislative language for a federal definition of SI, including an assessment of whether it should modify, expand upon or supersede the existing statutory text. The order also creates no right enforceable against the government, and it assigns the cost of its own publication to the Office of Science and Technology Policy.
Trump first floated the term at the United Nations General Assembly the week before. Speaking at the launch of America.gov on Tuesday, he told reporters that "super" was the best and simplest available word and said Chinese President Xi Jinping liked it too, according to The Verge. A White House fact sheet framed the rename as recognition that frontier systems do more than imitate human intelligence.
The four layers the labs agreed to
The accord, subtitled "Joint Commitment on Frontier Responsibilities," ran to roughly 300 words and was posted to Trump's Truth Social account. Its full text asks each signatory to monitor model capabilities and alignment during training and deployment in areas like cybersecurity, biosecurity and chemical threats; to empower an internal team to verify those controls work; to retain an independent external auditor; and to seat an independent board committee that receives reports from both.
That is the architecture of corporate financial oversight transposed onto model safety — controls, internal audit, external audit, board committee. What the document omits is every number that would make it testable. As The Register noted, there is no inspection frequency, no threshold for adequacy, and no external referee. Signatories will instead "meet regularly to establish standards and best practices" — writing the rules they are judged against. The text concedes that "over time, it may make sense to codify these steps into laws or regulations."
Trump called the accord morally binding and described the industry as self-policing. Musk, speaking later that afternoon, summarised it as joint monitoring, board special committees and generally grading each other's homework.
Why the administration prefers accords to agencies
Vice President JD Vance used his remarks to reject the idea of an FDA- or FAA-style regulator for frontier models, arguing that such a body would struggle to hire people who understood the technology as well as its builders. He told the assembled executives that taking the risks seriously themselves was preferable to a regulatory regime that could make matters worse, and said the Federal Trade Commission and the Justice Department already hold enough authority to pursue developers whose products harm consumers, Nextgov reported.
That position sits at the opposite end of the current congressional spectrum from the Ban Artificial Superintelligence Act, introduced six days earlier, which would license every frontier model through a new cabinet department and expose executives to 20-year sentences. Not all of the signatories arrive at the accord from the same place either: Zuckerberg and Huang have publicly opposed binding AI rules, while the others have argued for them.
What to watch next
The 60-day deadline for a proposed statutory definition of SI is the first hard checkpoint, and it is the point at which a rebranding either acquires legal content or stays cosmetic. The accord's own test is narrower: whether "meet regularly" produces a published standard with a cadence attached, or whether the four layers remain a description of what OpenAI, Anthropic and their peers already say they do. OpenAI had already shelved GPT-6.1 Astra for failing its internal alignment bar — evidence that company-run guardrails can stop a launch, and a reminder that the public hears about it only when the lab decides to speak.
FAQ
Does the executive order ban the term "artificial intelligence"?
Not in general. It applies to non-statutory documents produced by executive branch departments and agencies, and it does not require altering previously issued regulations, contracts, grants or historical records. Statutes passed by Congress still use "artificial intelligence," and the order's own definition of SI points back to that existing statutory language.
Is the White House Accord on Super Intelligence legally binding?
No. It is a voluntary commitment with no enforcement mechanism and no stated penalty. Trump described it as morally binding rather than legally so, and the document itself raises the possibility that the steps might be codified into law at some later point.
Which companies signed the accord?
Meta, Google, OpenAI, Anthropic, Nvidia and xAI. The signatures came from Google's Sundar Pichai, Anthropic's Dario Amodei, Meta's Mark Zuckerberg, OpenAI president Greg Brockman, xAI's Elon Musk and Nvidia's Jensen Huang, with Trump signing as well.
