White House 'Super Intelligence Accord' Is a Voluntary Pledge, Not AI Regulation

Executives from Google, Anthropic, Meta, OpenAI, xAI, and Nvidia signed "The White House Accord on Super Intelligence" on Tuesday, a voluntary agreement announced after a luncheon hosted by President Donald Trump that stops well short of the binding AI regulation many people inside those same companies have been asking for.

Trump called the effort an act of "tremendous self-regulation." The document itself uses softer language throughout. It says the companies "should implement" a set of safety practices — a formulation that leaves the commitments unenforceable as rules and enforceable, at best, as promises.

The accord lays out four obligations. First, companies "should implement" "robust internal controls" to monitor their models' capabilities and their ability to go rogue or hack systems in unintended ways. Second, they should "empower" an internal team to do that work and remediate problems when they occur. Third, the statement encourages the companies to partner with an external monitor who is "empowered" to conduct independent assessments verifying that the work is happening. Fourth, they should ensure their boards of directors have set up committees to receive reports on all of these efforts.

That structure — internal controls, an internal team, an external monitor, a board committee — mirrors the governance arrangements frontier labs have already been building on their own. What it does not do is create new legal obligations, independent auditing requirements with teeth, or penalties tied to specific failures.

A blessing, and an investigation

If the companies hoped Tuesday's luncheon would deliver a blanket blessing from the US government, that hope lasted less than a day. The New York Post reported that the Federal Trade Commission is planning a "sweeping probe" into Anthropic, OpenAI, other unnamed frontier labs, and METR, a nonprofit that conducts third-party evaluations of AI systems. METR and the labs did not respond to requests for comment.

An FTC spokesperson confirmed that some sort of investigation exists but declined to provide additional details, including which consumer protection issues concern the agency. According to the Post, the FTC has not yet issued civil investigative demands to its intended targets — usually one of the first formal steps in any investigation.

FTC chairman Andrew Ferguson, who appeared on the seating chart for the "super intelligence" luncheon, has previously described how enforcers begin without a complete case. "You're constantly looking for smoke to see if there's fire," he told a crowd of antitrust enthusiasts last year.

Google, Anthropic, Meta, OpenAI, xAI, and Nvidia did not respond to requests for comment. The White House also did not immediately respond.

What enforcement would actually look like

The voluntary commitments are some indication that frontier labs are trying to figure out how to make products safe for the general public. The stakes are considerable: researchers have warned that AI poses catastrophic risks if nothing is done, and the labs themselves have spent months debating what regulation should look like.

But the enforcement pathway for these particular pledges is narrow. Neil Chilson, a former chief technologist for the FTC, posted 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."

That enforcement runs through a specific legal theory: failing to honor a public promise can be considered a deceptive business practice under the FTC Act. The remedy, however, is usually modest. When the FTC alleges a company deceived its customers, the remedy it typically prescribes is making the company promise not to lie again.

A recent case illustrates the pattern. The FTC announced settlements with three companies it alleged were touting a service that could help businesses target ads based on audio recordings collected from smartphones, smart TVs, and other devices. The agency said the companies lied to businesses about their ability to creep on consumers. The fix: the companies are no longer allowed to misrepresent their advertising services and their collection and use of consumers' voice data. The case does not address whether collecting voice data from phones is permissible in the first place, and the settlement sets out no guardrails if one of the three companies wanted to do exactly that in the future — other than not lying about it.

Apply that template to frontier AI, and the ceiling becomes clear. A company that publicly committed to robust internal controls, ignored them, and got caught could plausibly end up signing a consent order requiring it to maintain robust internal controls. The accord alone does not trigger what anyone would call a dramatic reining-in of rogue "super intelligence."

There is precedent for federal interest in OpenAI specifically. The FTC reportedly sent a civil investigative demand to the company in 2023, though it is unclear what came of it. As experts have pointed out to WIRED, the US government can open investigations into companies even if enforcers eventually conclude there is no action to bring. OpenAI, for its part, has been remarkably open about the number of times its agents have broken into other people's systems and about its need to improve its safety practices.

Antitrust, extinction, and an odd exemption request

The statement caps weeks of debate over AI regulation, and it follows a quieter effort by frontier labs: seeking an antitrust exemption so they could coordinate on safety work without regulatory scrutiny. Antitrust experts and practitioners say such an exemption is unnecessary, on a pointed logical ground — AI ending humanity, which researchers have warned will happen if nothing is done, is arguably anti-competitive, because humans are an essential part of any given marketplace.

The labs' argument for coordination has some basis in practice. The UK and the Republic of Korea announced a set of "Frontier AI Safety Commitments" in early 2025, under which companies promised internal and external red-teaming of AI models for "severe and novel threats" and committed "to work toward information sharing" on AI safety issues. Tuesday's accord is not the first voluntary agreement these companies have signed with a government, and its structure suggests the same model: public commitments, government ceremonies, and enforcement only if a promise is broken.

The politics of enforcement

Even if the FTC investigation moves forward and the agency identifies violations, some observers doubt it will produce action against the AI firms.

"It should be encouraging to Americans worried about AI that the FTC is investigating these companies," says Douglas Farrar, who headed public affairs at the FTC during Lina Khan's tenure as chair. "But let's remember that Ferguson may chair the FTC, but Donald Trump runs it, and I doubt after yesterday's love fest with his AI CEO buddies that Trump will allow any legal action with teeth against these companies."

That political calculus matters more than the accord's text. The agreement binds the companies to nothing except their own credibility, and the FTC's strongest available remedy — a deception order — would leave the underlying safety questions unresolved. The real tests will come from what the FTC's investigation finds, whether civil investigative demands actually issue, and whether the external monitors and board committees the companies "should" establish ever publish anything the public can verify.

(Disclosure: This story's author previously worked in the Office of Technology at the FTC and resigned in November 2025.)

For now, the US approach to frontier AI safety rests on pinky-swears with a legal footnote: break a public promise and the government can make you promise again.