White House Accord on Super Intelligence: Signers and Terms

Four layers of controls and audits

The accord, subtitled “Joint Commitment on Frontier Responsibilities,” is one page. It opens with the premise that “every company is responsible for developing its own technology safely,” starting with “every company that is training and deploying frontier models” having internal processes to make sure its technology behaves as intended. It then says each company “should implement” four layers. [1][3]

First, internal controls that monitor models’ capabilities and alignment during training and deployment, in areas such as cybersecurity, biosecurity and chemical threats, and that ensure models “do not hack or access technical systems in unintended ways.” Second, an internal team empowered to make sure those controls work and that issues are fixed. Third, an independent external auditor or evaluator to assess whether the controls are working. Fourth, an independent committee of the board of directors that receives reports from the teams and auditors and makes sure problems are remediated. [3]

The participating companies also say they will “meet regularly to establish standards and best practices.” The closing paragraph allows that “over time, it may make sense to codify these steps into laws or regulations,” and says each company is committed to the controls regardless. [3]

Seven signatures, posted on Truth Social

Trump posted the accord as two images on Truth Social at 5:24 p.m. Eastern on Tuesday, September 29, after a White House lunch with AI executives. Beside his own signature are those of Sundar Pichai of Google, Dario Amodei of Anthropic, Mark Zuckerberg of Meta, Greg Brockman of OpenAI, Elon Musk, signing for xAI, and Jensen Huang of Nvidia. OpenAI was represented by its president rather than its chief executive, Sam Altman. [1][2]

The meeting was larger than the signature page. CNN and ABC News reported that Palantir’s Alex Karp and Amazon’s Jeff Bezos also attended, along with members of the Cabinet and House Speaker Mike Johnson; neither Palantir nor Amazon appears among the signatories. Pichai said on X that the accord “contains real tangible steps to promote safe development.” [2][5][4]

“Morally binding”

Asked whether the agreement binds anyone, Trump said it is “morally” binding and that there would be “a tremendous self-policing aspect.” He called it a “constitution.” Johnson called it “a statement of principles” and “commitments that are voluntary on behalf of the industry.” [5][4]

The text bears that out. It sets no penalties, no deadlines and no definition of a frontier model, names no auditor, does not require companies to publish audit results, and gives the government no role in enforcement. Each company decides how to implement the four layers. As of September 30 it was a signed statement of intent; no company had reported carrying it out. [3][4]

Atlas interpretation: The form is familiar. Sixteen companies signed voluntary frontier safety commitments in Seoul in 2024. The differences here are a presidential signature, a short and specific list of controls, and the explicit addition of an outside auditor and a board committee. Whether those become practice depends on what the companies publish next, which the accord does not require. [3][4]

From “hoax” to a shared document

Two weeks earlier, Dario Amodei had called on AI labs to pace the frontier. Fortune reported that Trump answered on Truth Social by calling the existential threat a “hoax” and naming Amodei, and at the UN on September 22 he rejected international AI oversight as a “globalist scheme.” [7]

On the night of Sunday, September 27, Trump and Amodei had a private dinner at the White House, their first one-on-one meeting. Before it, Trump said they would talk about AI regulation and that he was “for let’s go and let’s win.” The White House released no account of what was discussed. Two days later Amodei signed the accord. [6][5]

Atlas interpretation: The dinner is context for this page rather than an Atlas event of its own: it produced no announced decision, and the commitment that followed is recorded here. The same afternoon, Trump signed an executive order renaming AI “Super Intelligence” across federal agencies, which explains the accord’s title. [6][2]