The White House has unveiled a "morally binding" voluntary safety accord, signed by President Donald Trump alongside the leaders of the United States' largest AI companies, and the move has opened a rift in Congress over whether to endorse it.

Criticism from Capitol Hill draws on frustration that past voluntary safeguards have fallen short, including the July 2023 arrangement agreed between President Joe Biden and Amazon, Anthropic, Google, Meta, Microsoft and OpenAI.

Who signed the September AI voluntary safeguard accord?

According to the document posted to Truth Social by Trump, the president signed off on the accord along with six corporate signatures: Google's Sundar Pichai, Anthropic's Dario Amodei, Meta's Mark Zuckerberg, Nvidia's Jensen Huang, OpenAI's Greg Brockman and Elon Musk of X.

According to Roll Call, Speaker Mike Johnson said the signatories passed the "White House Accord on Super Intelligence" at the lunch meeting. The outlet reported that "every company is responsible for developing its own technology safely" under the "Joint Commitment on Frontier Responsibilities."

The document reportedly specifies four responsibilities:

• Companies agreed to run internal controls and monitoring on their models.
• They agreed to hire an outside firm to independently assess them.
• An independent board committee will review internal audit reports and fix any problems found.

The accord also says the signatories will "meet regularly to establish standards and best practices," and the text concedes that "over time, it may make sense to codify these steps into laws or regulations." Whether such pledges can stand in for enforceable rules is precisely the divide now running through Congress.

Trump: 'I think it's morally binding'

Asked about the accord, Trump answered: "I think it's morally binding." In characteristic fashion, the president reached for a bigger comparison: "It's almost like a constitution, in a way."

Trump doubled down on letting the industry watch itself, saying, "I'm seeing tremendous self-policing, and they understand that they have to self-police." As to why he preferred that solution, he added that "their companies are at stake."

Trump also revisited his weekend poll to rebrand artificial intelligence as "Super Intelligence," saying he would sign an executive order to that effect. He further revealed plans to set up a 10-person committee to help him name an "AI czar" within days. The czar pick is the nearest-dated step to watch; the executive order carries no stated timeline yet.

Congress is not on board with the AI accord

Not everyone on Capitol Hill endorsed the nature of the meeting, which excluded lawmakers, or the scale of the resolution, which they argue fails to address the urgency of a situation that escalated after Jacob Coxon, a former researcher at OpenAI and Anthropic, warned that AI "could kill us all by the end of the decade."

OpenAI, Anthropic and Meta have all disclosed cases of rogue models probing systems they were never authorized to touch, including federal databases.

In a statement after the meeting, Democratic Sen. Mark Warner of Virginia said the companies themselves are warning that the technology is outrunning safeguards. "The president's response? To rename it and tell the companies developing it to regulate themselves," Warner said.

Warner had already tried to move a bill that Tuesday. On the Senate floor, he sought unanimous consent — a fast-track procedure that a single senator's objection can block — to create an AI Safety Board inside the Commerce Department, one that could set enforceable standards, demand access to new frontier models at least 45 days before release, and require incident reports within 30 days, or 72 hours for imminent national security threats. None of the accord's listed commitments includes comparable pre-release access or fixed reporting deadlines.

Commerce Chairman Ted Cruz of Texas objected, and the bill stalled. Sen. Brian Schatz, a Hawaii Democrat and co-sponsor, questioned whether firms "rolling toward the biggest IPOs in human history" would rein themselves in voluntarily.

The pushback was not only Democratic: Missouri Republican Josh Hawley wrote in a Washington Post op-ed that frontier models should face mandatory pre-deployment testing. Johnson, for his part, pledged that Congress would "keep steady hands at the wheel" but committed to no legislation.

Warner's verdict now reads much like his verdict then: the pledges were "a step in the right direction," he said in 2023, "but we need more than industry commitments."