Trump's White House AI Accord Leaves Safety to Companies Themselves

President Donald Trump hosted the CEOs of Alphabet, Meta, OpenAI, Anthropic, Nvidia, Palantir and SpaceX at the White House on Tuesday and got them to sign a two-page voluntary document that keeps AI safety oversight almost entirely in the industry's own hands.

The document, titled "White House Accord on Super Intelligence: Joint Commitment on Frontier Responsibilities," commits each company to developing its own technology safely — but creates no regulator, no enforcement mechanism and no external standard. Trump posted the signed accord on his Truth Social account.

The stakes are unusually high. The lunch came at the end of a three-week stretch in which AI safety became an intense nationwide debate, with leading industry researchers and executives publicly warning of extreme threats posed by the most powerful models. Anthropic CEO Dario Amodei published an essay earlier this month urging AI labs to "pace the frontier." Yet the same executives who warned their technology cannot be trusted in the absence of coordinated safety efforts have now signed a document saying "every company is responsible for developing its own technology safely and in a way that builds trust with customers and the public."

Trump has steadfastly opposed government regulation of AI throughout this debate and has called AI fears a "hoax."

Four voluntary action items

The accord lays out four commitments. Each company agrees to establish "robust" internal monitoring of its models, maintain an internal team that ensures the controls are working, partner with outside auditors or evaluators, and designate an independent committee of the board to oversee that internal team.

The document's final paragraph gestures at future rules: "Over time, it may make sense to codify these steps into laws or regulations."

When reporters asked Trump outside the White House whether the rules are binding, he answered that they are "morally binding."

Industry watchers immediately flagged the tension between the CEOs' earlier positions and Tuesday's signing. "You saw two weeks ago almost all those same CEOs say they should be regulated," Bradley Tusk, CEO of Tusk Ventures, said in an interview on CNBC's "Closing Bell Overtime" on Tuesday. "It's not because they fundamentally believe in the power of the state. They want something that creates an equalizer for each other."

A rebrand, too

The accord was not the only item on the agenda. Trump separately said he signed a document purporting to "officially" change the name of AI to "super intelligence," pushing through a rebrand after crowdsourcing a new name on social media last week. Tech companies are unlikely to abandon the AI acronym anytime soon.

The executives, for their part, were quick to voice support for the broader effort — a consistent posture during Trump's second term as corporate America, and particularly the tech industry, works to stay out of the president's crosshairs.

Meta CEO Mark Zuckerberg told reporters the agreement is designed to "give the American people and our customers confidence that the technology works in the way we intend."

"The idea isn't that this is the only thing that we'll ever do," Zuckerberg said. "It is that this is a start and an accord that the whole industry can come to."

Zuckerberg has sat closer to the self-regulation camp than peers like Amodei and OpenAI CEO Sam Altman, alongside Nvidia CEO Jensen Huang. Days after Amodei's essay, Zuckerberg wrote in a post on X that "trust and alignment are quickly becoming the most important capabilities that will differentiate agents and models."

Amodei, who has had a rocky relationship with the White House but had dinner with Trump late last week, offered a measured comment to reporters after the signing: "If we do this right, if we work with the president and everyone here we can win safely."

Alphabet's Sundar Pichai called the moment "historic and consequential."

Pressure for regulation is building anyway

Public sentiment toward AI companies has turned decidedly sour heading into the November midterm elections, and concerns about high-profile security breaches continue to escalate. The political pressure for binding rules is likely to grow, not shrink.

The numbers back that up. According to Verasight's September national survey, 63% of respondents said AI development should slow down, while only 5% said it should accelerate.

Lawmakers are already moving. Sen. Mark Warner, D-Va., said Tuesday that he has introduced legislation to require "rigorous testing and evaluation" of leading AI technology and to strengthen security and reporting "before the most powerful models are deployed."

"The companies building the most powerful AI systems are warning us that the technology is advancing faster than our safeguards," Warner said in a release. "The president's response? To rename it and tell the companies developing it to regulate themselves. But changing what we call artificial intelligence does nothing to address the very real risks that come with increasingly powerful AI systems."

CNBC's Kevin Breuninger, Samantha Subin and Lora Kolodny contributed to the reporting on which this article is based.

With the industry policing itself, a voluntary accord in place of statute, and Congress now circling with binding legislation, the gap between what AI executives say about their own risks and what they have formally committed to do about them is set to define the policy fight through November and beyond.