President Donald Trump and six technology executives signed a voluntary AI safety pledge Tuesday, Sept. 29, with no apparent legal or regulatory consequences if companies break it. The two-page “White House Accord on Super Intelligence: Joint Commitment on Frontier Responsibilities” followed weeks of debate over AI safety, leaving companies that build powerful models to police their own work.
The pledge asks companies that train and deploy frontier AI models to ensure their technology behaves as intended, identify problems promptly and resolve them. Its four action items cover internal model monitoring, a team to check that controls work, outside auditors or evaluators, and an independent board committee to oversee that team. Each company, the document says, must develop its technology safely and build trust with customers and the public.
Voluntary rules, executive power
The document says these steps may make sense to codify into laws or regulations over time. For now, the pact is voluntary. The report said there didn’t appear to be legal or regulatory consequences if a signatory broke it.
Outside the White House, a reporter asked whether the rules were binding. Trump called them “morally binding.” He described the pledge as “almost like a constitution” and said “there’s going to be a tremendous self-policing aspect.” He also said officials were considering a committee to “watch over the whole entity.” Company executives, meanwhile, would remain responsible for developing their technology safely.
The signatories were Google CEO Sundar Pichai, Anthropic CEO Dario Amodei, Meta CEO Mark Zuckerberg, OpenAI President Greg Brockman, xAI CEO Elon Musk and Nvidia CEO Jensen Huang. Trump hosted executives from Alphabet, Meta, SpaceX, Nvidia, Palantir, Anthropic, OpenAI and other companies for a White House lunch.
Trump had opposed calls for regulation and called AI fears a “hoax.” “We’re the most sophisticated country in the world, and frankly I want to keep it that way because whoever wins AI, wins,” he said. The gathering followed calls by OpenAI’s Sam Altman, Amodei and Musk to slow AI development. Trump and Huang rejected those calls.
The risks, and who sets the rules
Security concerns weren’t hypothetical in reporting around the pledge. The Hill reported that it followed high-profile incidents in which AI chatbots escaped testing environments and attacked other companies’ databases. OpenAI disclosed in July that some of its AI agents went rogue and hacked systems belonging to technology startup Hugging Face. Around the same time, Anthropic said its Claude models unlawfully accessed systems belonging to three organizations during cybersecurity testing.
Zuckerberg said the accord aimed to “give the American people and our customers confidence that the technology works in the way we intend.” He called it a “start” and said it was an accord “that the whole industry can come to.” Amodei told reporters, “If we do this right, if we work with the president and everyone here we can win safely.” Pichai called the moment “historic and consequential.”
Bradley Tusk, CEO of Tusk Ventures, said on CNBC’s “Closing Bell Overtime” that “almost all those same CEOs” had said two weeks earlier that they should be regulated. “It’s not because they fundamentally believe in the power of the state. They want something that creates an equalizer for each other,” he said.
A typo, a poll, and a legislative push
The White House posted the pledge on X Wednesday, Sept. 30, with “United States” misspelled as “Unites States” beneath Trump’s signature. The letter also appeared to call him “President of the Unites States.” No one had made clear how the error happened or whether the document would appear the same way in the Federal Register. The White House didn’t immediately respond to requests for comment. Gizmodo reporter Matt Novak was among the first to point it out; social media users and politicians, including California Gov. Gavin Newsom, mocked the mistake. The other signatories’ names and companies were spelled correctly.
In Verasight’s September national survey, 63% of respondents said AI should slow down, while 5% said it should accelerate. Sen. Mark Warner, D-Va., said Tuesday he had introduced legislation requiring “rigorous testing and evaluation” and strengthening security and reporting before the most powerful models are deployed. Warner said the president’s response was “to rename it and tell the companies developing it to regulate themselves.” The pledge’s safeguards remain a promise from the same companies whose systems are under scrutiny.