Trump got AI’s biggest bosses to sign a “morally binding” pact, and it comes with no new rules

President Trump says the heads of America’s biggest AI companies have signed a pact to police themselves, and that it’s all the regulation they need. After a White House lunch on Tuesday with Elon Musk, Mark Zuckerberg, Dario Amodei, Jensen Huang, Sundar Pichai, OpenAI’s Greg Brockman and others, Trump told reporters the executives had signed an agreement that is “morally binding”, CBS News reported.

“It’s almost like a constitution”

House Speaker Mike Johnson, who has pushed for the meeting for months, described the document as a “joint commitment” and a “statement of principles” on what Trump now calls “super intelligence”. It commits the companies, voluntarily, to “robust internal controls and layers of internal and external review,” he said, and its text will be released later.

Asked whether the deal was binding, Trump said:

I think it’s morally binding. It’s almost like a constitution, in a way.

Donald Trump, US president

Zuckerberg told reporters the firms had agreed to “multiple layers of auditing and controls”, including internal evaluations and outside audits, with each company’s board reviewing the auditors’ findings. Trump also floated a committee of about 10 people to “watch over the whole enterprise”, and said he would name an “AI czar” within days.

“We automatically have regulation”

What Trump ruled out was new law. According to Bloomberg’s account, he said existing agencies were enough:

There’s a belief that there should be tremendous self-regulation, and we automatically have regulation with the Department of Justice, the FBI, all of that, but the self-regulation is very important.

Donald Trump, speaking outside the West Wing

“I think I’m seeing tremendous self-policing. And they understand that they have to self-police,” he added. Trump also said the companies would do more for communities near new data centres, without giving details.

A turn, of sorts

Just a week ago, Trump was calling fears of catastrophic AI risk a “hoax”. Holding a summit on AI safety at all marks a shift, after a month of rogue AI incidents, OpenAI cancelling its next model and polls showing voters increasingly worried ahead of November’s midterms. Amodei, who has called for regulation, sat down with Trump for a private dinner on Sunday.

But a voluntary pledge is a long way from what critics want. Bill Gates warned last week that self-regulation isn’t enough, and Congressional Democrats including Alexandria Ocasio-Cortez are pushing a bill to ban superintelligence outright. Johnson has said Congress may need to act, but that it would take time.

Update: the “Super Intelligence” order is signed

Later on Tuesday, Trump signed an executive order telling federal agencies to say “Super Intelligence” and “SI” instead of “AI”. Here’s what it does. The accord’s text and the names of the proposed committee and AI czar still haven’t been published.

Update: what the accord actually says

The text has now been published in full. Titled the “White House Accord on Super Intelligence: Joint Commitment on Frontier Responsibilities”, it is signed on behalf of Google, Meta, Anthropic, SpaceXAI, Nvidia and OpenAI.

It sets out four “layers” of controls for companies building frontier models: internal monitoring for cyber, biological and chemical risks, including making sure models don’t “hack or access technical systems in unintended ways”; an internal team to check those controls work; independent external auditors; and a committee of each company’s board to review the findings. The companies also promise to meet regularly to agree safety standards.

There are no deadlines, no penalties and no role for a regulator. But the document goes a step further than Trump did at lunch: it says that over time it “may make sense to codify” some of these steps “into laws or regulations”.

Update: behind closed doors, the other bosses questioned Amodei

The show of unity at lunch hid a row. After the meal, in a smaller session in the Roosevelt Room, Nvidia’s Jensen Huang and other executives asked Amodei why he was so extreme in public about AI’s risks, The Wall Street Journal reported on Wednesday, citing people familiar with the meeting. Amodei replied that it was important to be honest about the risks, including cyberattacks and job losses.

The Journal also reported that an earlier plan for an industry self-regulatory body backed by Anthropic, OpenAI and Google was dropped after Huang, Meta’s Mark Zuckerberg and Elon Musk objected that it would hand too much power to the leading labs. Anthropic and Nvidia haven’t commented on the report.

Update: the Senate’s Republican leader wants it in law

Senate Majority Leader John Thune welcomed the accord but said he wants to “codify” some of its protections. “We are continuing to have those conversations up here on whether or not we can come up with some sort of a legislative framework that would codify some safety protections too,” he told Axios on Wednesday. He didn’t say when a bill might come, and Congress has left Washington for the midterm campaign. It comes a day after Senator Ted Cruz blocked a Democratic bill that would have made AI safety rules mandatory.

The White House has since put intelligence chief Jay Clayton in charge of AI, at the head of a “Super Intelligence Force” due to report on the risks in 120 days.

Why it matters

The companies building the most powerful AI have now made a public promise to audit themselves, and the White House has made clear that promise is the plan, not a stopgap before new law. Until the text is published, nobody outside the room knows exactly what was signed, or what happens if a company breaks it.

Sources: CBS News, Bloomberg via The Sydney Morning Herald (September 29, 2026), accord text via the Washington Examiner, The Wall Street Journal (September 30, 2026), Axios (October 1, 2026)