Is Mark Zuckerberg the new Trump whisperer?
At the White House state dinner for Chinese President Xi Jinping on Sept. 24, Mark Zuckerberg was seated beside House Speaker Mike Johnson. The Meta chief executive talked to the Louisiana Republican about regulatory concerns around AI, people familiar with the conversation said.
The idea for the accord President Donald Trump and six technology executives signed five days later, on Tuesday, came from that conversation. The text ran to about 300 words.
Zuckerberg, a recurring target of both parties at congressional hearings, originated and circulated the industry's latest self-regulation pledge. In August, he had opposed a standing body modeled on the Financial Industry Regulatory Authority to oversee AI companies, while the pledge leaves each company to apply controls to itself.
The Breakdown
* The idea for the White House Accord on Super Intelligence came from Mark Zuckerberg's conversation with House Speaker Mike Johnson at the Sept. 24 state dinner for Xi Jinping.
* Zuckerberg circulated a draft, Jensen Huang gathered support among the AI labs, and Trump signed the roughly 300-word accord with six executives on Sept. 29.
* In August, Zuckerberg opposed a standing body modeled on the Financial Industry Regulatory Authority; the accord leaves each company to apply four layers of controls to itself.
* All accounts of his role come from anonymous sources. Meta declined to comment, and Hakeem Jeffries called the agreement entirely unenforceable.
AI-generated summary, reviewed by an editor. More on our AI guidelines.
## The dinner
Semafor's exclusive account traced the proposal to the dinner conversation. Zuckerberg subsequently approached Nvidia chief executive Jensen Huang, who sat next to Trump at Tuesday's lunch, about principles the executives could take to their next meeting with Trump. Huang gathered support among the AI labs, and Zuckerberg circulated a draft before Tuesday's lunch.
Every account of Zuckerberg's role rests on people who spoke on condition of anonymity; Meta declined to comment, and the White House and Nvidia did not immediately respond.
As the text took shape, Zuckerberg frequently spoke with Trump and Commerce Secretary Howard Lutnick, who oversees a key government AI testing unit, people familiar with the discussions said. Huang and administration officials also helped shape the agreement.
Johnson had already set out his position. “We do not need a moratorium,” he said on Monday. “We do not need to jump in and hyperregulate this, because we'll lose the race to China.”
## The plan that was scrapped
During the summer, Anthropic, OpenAI and Google had coalesced around a different arrangement the administration was considering: a self-regulatory organization modeled on the Financial Industry Regulatory Authority, which oversees brokerage firms. The proposal also included stronger internal reviews during model development and external testing.
FREE WEEKDAY MORNING BRIEFING
Follow who is writing the rules for AI.
The Implicator Morning Briefing filters the AI news cycle to the stories worth your attention and explains their consequences. From San Francisco, every weekday at 4:45 a.m. Pacific, 7:45 a.m. Eastern.
Email address
Send me tomorrow’s briefing
Check your inbox for the confirmation link.
About five minutes. No hype. No spam.
Huang, Zuckerberg and Elon Musk told Trump they worried that arrangement would concentrate too much power in the leading AI companies, people familiar with the talks said. The idea was scrapped. Zuckerberg opposed it directly in an August phone call with Trump.
The September agreement retained internal reviews and outside evaluation. Its four voluntary layers consist of internal controls to monitor models during training and use, including cybersecurity and biological or chemical risks; an internal team to check the controls and fixes; an outside auditor or evaluator to assess them; and an independent board committee to receive reports and oversee corrective action. The text does not say who carries out outside evaluations, who appoints or pays the auditor, or whether findings will be published.
Trump signed the two-page accord on Sept. 29 alongside Google's Sundar Pichai, Anthropic's Dario Amodei, Zuckerberg, OpenAI president Greg Brockman, xAI's Musk and Huang. The companies also pledged to meet regularly to establish safety standards.
Its final paragraph states: “Over time, it may make sense to codify these steps into laws or regulations.”
Meta had signed voluntary safety commitments with President Joe Biden in July 2023. Trump revoked the executive order that followed those commitments on his first day back in office in January 2025.
## The lunch
Amodei had dined with Trump on Sunday, their first face-to-face meeting. At Tuesday's lunch, he raised safety concerns. Zuckerberg answered that the best way to address them was for the industry to follow through on the principles, people familiar with the exchange said.
Later, in the Roosevelt Room, a smaller group of executives and White House advisers worked on the principles. Other executives, including Huang, asked Amodei why he was being so extreme in public about AI capabilities and risks. Amodei replied that being honest with the public mattered and that the dangers should not be played down, people familiar with the discussion said.
Outside, a reporter called Amodei forward to answer a question. He spoke about winning safely with the president and the industry, then retreated to the back. Zuckerberg was smiling and laughing with Trump as they took questions.
Know someone who'd find this useful? ✉️ Email it to a friend in one click, or they can subscribe free here.
His company had settled Trump's lawsuit over the suspension of his social media accounts after the Capitol attack. During Trump's second term, Meta also rolled back fact-checking, hired executives friendly to the president and donated to his causes.
Trump called Tuesday's agreement “morally binding.” Asked why the executives should be trusted to regulate themselves, he answered: “Because they're outstanding people.”
Zuckerberg described the agreement as a beginning. “The idea isn't that this is the only thing that we will ever do. It is that this is a start and an accord that the whole industry could come to.”
## The critics
Outside the Capitol on Wednesday, House Minority Leader Hakeem Jeffries called for Congress to act immediately. He pointed to episodes in which AI agents had operated outside their parent companies' instructions without detection and called for action “not in the next Congress, but right now.”
“The notion of entering into a voluntary agreement that is entirely unenforceable, and that allows the industry writ large to police itself, is not the right response at this moment given the potentially grave harm and danger that industry leaders themselves are now flagging for the American people,” Jeffries said.
Alvin Wang Graylin, a senior fellow at the Asia Society Policy Institute's Center for China Analysis, saw value in getting the rivals to commit. “Putting that on paper, with Zuckerberg and Huang in the room alongside Amodei, is better than another year of 'we cannot regulate because China will not'.”
Graylin said: “What is missing is independence and anything that crosses a border. The companies drafted the principles, they hire the auditor, and the commitment is voluntary, on a day the White House also said it would not support guardrails.”
Frequently Asked Questions
How did Mark Zuckerberg shape the White House AI accord?
The idea came from his conversation with House Speaker Mike Johnson at the Sept. 24 state dinner for Chinese President Xi Jinping. Zuckerberg then approached Nvidia's Jensen Huang about a set of principles, circulated a draft before the Sept. 29 White House lunch, and spoke frequently with Trump and Commerce Secretary Howard Lutnick as the text took shape.
What does the White House Accord on Super Intelligence require?
The roughly 300-word, two-page accord asks each company to adopt four voluntary layers: internal controls to monitor models, an internal team to check them, an outside auditor or evaluator, and an independent board committee. It does not say who appoints or pays the auditor, or whether findings will be published.
Who signed the accord?
President Trump signed it on Sept. 29 with Google's Sundar Pichai, Anthropic's Dario Amodei, Meta's Mark Zuckerberg, OpenAI president Greg Brockman, xAI's Elon Musk and Nvidia's Jensen Huang. Trump called it "morally binding."
What alternative did Zuckerberg oppose?
Over the summer, Anthropic, OpenAI and Google coalesced around a self-regulatory organization modeled on the Financial Industry Regulatory Authority. Zuckerberg opposed it in an August phone call with Trump, and he, Huang and Musk said it would concentrate too much power in the leading AI companies. The idea was scrapped.
How have critics responded?
House Minority Leader Hakeem Jeffries called the voluntary agreement entirely unenforceable. Alvin Wang Graylin of the Asia Society Policy Institute said putting it on paper beat another year of “we cannot regulate because China will not,” but noted the companies drafted the principles and hire the auditor.
AI-generated summary, reviewed by an editor. More on our AI guidelines.
Six AI Chiefs Sign a Pledge With No Penalty as Washington Declines to RegulateOn Tuesday, September 29, 2026, President Donald Trump came outside the West Wing after lunch with technology executives. Meta chief executive Mark Zuckerberg stood with the group as reporters asked aThe ImplicatorTrump Orders Agencies to Call AI 'Super Intelligence' but Keeps the Legal DefinitionPresident Donald Trump signed an order Tuesday requiring federal agencies to call AI “Super Intelligence.” The order defines the new term as the same technologies covered by the existing statutory defThe Implicator