President Donald Trump and six technology executives signed a one-page voluntary agreement on Tuesday that asks companies to police their own artificial intelligence systems, an approach Trump has favored over federal regulation. Trump posted the document on Truth Social, calling the commitment morally binding and relying on reputation rather than law to enforce it. The signatories were Anthropic's Dario Amodei, Google's Sundar Pichai, Meta's Mark Zuckerberg, OpenAI President Greg Brockman, Nvidia's Jensen Huang, and xAI founder Elon Musk.
Four layers of self-policing
The accord sets up what amounts to a four-step oversight chain inside each participating company: internal controls, an internal verification team, an outside auditor, and an independent board committee. In effect, the companies would check their own work at each stage. The pledge remains voluntary, and the reporting does not say who selects the auditors, which standards they will apply, or what happens when a control fails. Those blanks are significant because the agreement is only as strong as the details it leaves unaddressed.
Trump keeps the door open on an overseer
Trump said he will name an overseer in the coming days and suggested roughly 10 people would sit on a committee. That structure is not yet public, and no names have been released. The White House has not described how the committee would interact with the four-layer framework inside each company. For now, the accord is a document on Truth Social and a set of signatures, not a regulation with penalties attached.
A race with China, and no room for second
Trump framed AI on Tuesday as a race with China that leaves no room for second place. That framing has shaped his approach: rather than impose federal guardrails, he has pushed companies to set their own rules. The accord reflects that position. It also arrives as data center opposition is growing in some communities and as AI agents have gone rogue in several high-profile incidents, raising questions about what happens when automated systems misbehave.
Signatories with different views
The signatories are not of one mind. Dario Amodei of Anthropic urged a slowdown in an essay earlier this month, saying the technology carries real risks. His company signed anyway. House Speaker Mike Johnson has argued that AI safety falls on developers, a view that lines up with the accord's structure. The mix suggests the agreement is less a consensus on safety than a shared willingness to be seen committing to something, with each company interpreting the terms its own way.
What the accord doesn't say
The one-page document leaves many operational questions open. Who picks the outside auditors? What standards do they use? What triggers a review? What happens if an internal control fails or an independent board committee overrides a company decision? None of those answers are in the facts released so far. Trump said the commitment is morally binding, which means the consequences for a violation would come through reputation, not courts or agencies. That puts pressure on the overseer and the committee he plans to name, since they would be the only visible check on companies that have agreed to check themselves.
The next concrete step is Trump's announcement of an overseer in the coming days, followed by his suggestion of a roughly 10-person committee. Until those names and that structure are public, the accord's four layers exist mostly on paper.
