Science1 publisher2 min readPublished
Trump offers a voluntary AI review accord in place of new federal development limits
Donald Trump said AI leaders signed a voluntary review accord on September 29, though none of the eight executives at the meeting is a confirmed signer. As he described it, oversight rests on company self-policing and existing DOJ and FBI powers.
The Scientist · Science desk
Drafted by a language model from the sources cited here and checked against its claim ledger before publication. How we use AISend a correction
What happened
- Trump rejected new federal limits on AI development the same day, promising the government would not back guardrails or other restrictions.
- Attendees included Amodei, Brockman, Bezos, Huang, Musk, Pichai, Zuckerberg and Nadella, joined by JD Vance, Scott Bessent and Mike Johnson.
- The companies represented are spending hundreds of billions of dollars on AI infrastructure, and their data center projects have drawn grassroots opposition across party lines.
- Trump said the executives agreed to answer that local opposition with more money for local schools or with steps to cut energy costs.
Compiled by The ScientistSomething wrong?How this is made
Why it matters
- constraint Without a published test or a named reviewer, people outside the companies cannot check whether a review was run or what it found.
- decision Each company decides for itself whether to act on Amodei's call to slow down, because the only federal check Trump named is authority the Justice Department and FBI already held.
- cost Towns that resist data centers are asked to weigh school or energy help of no stated size against Trump's warning that the projects could move overseas.
Trump said the accord would include both internal and external reviews of AI [1]. An external review tells people outside a company something only when the test and the tester are fixed before any results come in, the way a clinical trial registers its endpoints first. The Gulf Coast News report on the accord, as summarized by Superpower Daily, specified neither. It gave no review standards and did not identify who would conduct the external reviews [4].
Trump's own description put the weight on the companies. "I think I'm seeing tremendous self-policing. And they understand that they have to self-police," he told reporters outside the West Wing [6]. After the lunch he said there was widespread agreement on self-regulation, "and we automatically have regulation with the Department of Justice, the FBI, all of that" [7]. Earlier the same day he had said, "I will never stifle the growth of a technology that will be bigger than the industrial revolution" [3].
The executives in the room did not come with one view of safety. Anthropic's Dario Amodei had called earlier in September for leading firms to slow development. Nvidia's Jensen Huang and Meta's Mark Zuckerberg had played down concerns about runaway AI and stressed each company's responsibility for product safety [8]. I think the accord, as Trump described it, is the Huang and Zuckerberg position announced from the White House: safety as each company's own job [c6, c8]. That judgement holds only for the version Trump gave reporters. A published text naming reviewers and standards would change it.
The whole record is one news report, relayed secondhand, plus the president's remarks [c4, c6]. Superpower Daily's summary called the practical terms of the accord unclear [14].
Trump also predicted that local opinion on data centers would turn. "You're going to see, data centers are going to be very popular," he said [12].
What to watch
- Release of the accord's text or a signer list, including who performs the external reviews.
- Any company naming an outside reviewer or publishing the standard it will review a model release against.
- Dollar figures or signed local agreements for school funding or energy-cost relief near data center sites.