4 wild moments from Trump’s AI safety lunch with tech leaders

President Donald Trump’s White House lunch with some of the biggest names in tech came with some big moments.

It brought together a who’s who of tech leaders — from Elon Musk and Nvidia CEO Jensen Huang to Meta CEO Mark Zuckerberg and Anthropic CEO Dario Amodei — to discuss the future of AI and how to keep it safe.

It produced a voluntary agreement, an interesting exchange with Amodei, and a typo that critics quickly seized on.

Here are four of the most eye-catching moments.

Trump said executives, including Google’s Sundar Pichai and OpenAI president Greg Brockman, signed what he described as a kind of “constitution” for the industry.

When asked by a reporter if the accord was binding, Trump offered a qualifier: “I think it’s morally binding.” The one-page White House Accord on Super Intelligence asks companies to build internal safety controls, have external auditors assess them, and submit the resulting reports for independent board review.

It contains no legal enforcement mechanism.

Trump also said the companies would engage in “tremendous self-policing” and floated the idea of a committee of roughly 10 people to “watch over the whole enterprise.” One of the most talked-about clips from the post-lunch press gaggle came when Trump called Amodei forward.

Trump had previously criticized the Anthropic CEO over his calls for stronger AI safeguards and a slower pace of frontier development.

On Tuesday, he praised Amodei, saying he had been “fantastic.” Then, as Amodei stepped forward, Trump told the press, “Whatever he says is OK,” before giving the CEO some advice: “Be careful.” Trump does this better than most people.A light moment in a very serious lineup: Dario Amodei with President Trump in front of the press.“whatever he says is okay.” Then: “Be careful.” pic.twitter.com/jfY5hJ7Dfo Amodei told the gathered media that AI had “incredible benefits” but “very real risks,” and said companies and the government needed to work together so the US could “win safely.” As Amodei spoke, Trump and Zuckerberg stood behind him and shared a whispered exchange, followed by laughter.

During that same press conference, Zuckerberg later explained that the accord was a starting point built around internal risk reviews, external audits, and board oversight.

Trump shared the signed document on Truth Social, calling it the “White House Accord on Super Intelligence.” But eagle-eyed critics noticed a spelling error beneath Trump’s signature: “President of the Unites States.” Superintelligence community has a spelling error 🤣 pic.twitter.com/JAkN70DUpu The internet was quick to seize on the mistake in a document that the president had likened to a constitution.

A close-up of the tech leaders’ signatures showed most looking like conventional autographs.

OpenAI’s Brockman stood out for its simplicity: it was simply the lowercase letters “gdb.” Musk appeared to get tripped up by Trump’s new terminology for AI — “Super Intelligence” — during the press gaggle. “I think it is worth highlighting the positive benefits of AI — S.I., pardon me,” Musk said, quickly correcting himself while standing next to Trump.

Musk then made the case that the technology could help usher in an era of abundance.

The slip came hours after Trump signed an executive order directing federal agencies to replace “Artificial Intelligence” and “AI” with “Super Intelligence” and “SI” in official communications, wherever legally permitted.

Trump said he disliked “artificial” because it sounded “fake.”