Buyers should file the White House AI accord as a vendor promise the FTC might police

Sometime this week, a model provider's security questionnaire will probably gain a new line under governance: signatory, White House Accord on Super Intelligence ^[1]. The reviewer then has to decide which box that line fills.

The accord lists four things companies "should implement" ^[2]. They are "robust internal controls" to monitor what their models can do, an empowered internal team to run that work and fix failures, an external monitor to make independent assessments, and a board committee that receives reports on all of it ^[2]. On the monitor, WIRED reports, the text only encourages labs to partner with one ^[2]. Three of the four provisions are carried out by the lab's own staff or its own board ^[1].

WIRED's assessment is that the accord falls short of the sweeping regulation that people inside the companies have themselves asked for ^[12]. The enforcement route WIRED identifies is consumer-deception law ^[8]. Neil Chilson, a former chief technologist at the FTC, wrote on X that "this is the kind of pledge that the FTC could potentially enforce, if a company materially failed to follow through on any of these promises" ^[4].

Under the FTC Act, failing to keep a public promise could be considered a deceptive business practice ^[8]. WIRED's reporter disclosed working in the FTC's Office of Technology until resigning in November 2025 ^[13]. According to WIRED, the agency's usual remedy for deception is making the company promise not to lie again ^[9]. Its example is a recent settlement with three companies the FTC said had lied to businesses about an ad-targeting service built on audio from phones, smart TVs and other devices ^[10]. The order bars them from misrepresenting the service and their use of voice data ^[10]. It does not decide whether collecting that audio is acceptable, and it sets no guardrail on doing it later beyond telling the truth about it ^[11].

For a buyer, I'd sort each of the four provisions on two questions: can anyone outside the lab see evidence that it is happening, and is it written into your agreement with the vendor. A provision that is visible and in the contract is a control you can enforce yourself. In the contract but unseen, you are relying on attestations, and the board committee's reports are the obvious thing to request. Visible but outside the contract gives you material for an FTC complaint and no leverage with the vendor. Every provision starts outside the contract column, because the labs made these commitments in a public statement at the White House ^[1][8].

The external monitor is the provision to press on. It is the only one of the four that brings in a party outside the company ^[1]. I'd accept the accord line in a questionnaire as a statement of intent, then ask each vendor to name its monitor and share the assessments under contract. The cost of that approach is negotiating leverage. A lab that signed a "should" may decline to put a "must" in one customer's paper, and smaller buyers have less room to insist.