Mark Zuckerberg circulated the draft of the AI safety pledge Donald Trump released Tuesday as the "White House Accord on Super Intelligence," Semafor reported Wednesday, citing a source familiar with the matter and two others who confirmed his role. Meta and the White House declined to comment.
The idea started at a dinner table. Zuckerberg sat next to House Speaker Mike Johnson at last week's state dinner for Xi Jinping and raised his worries about AI regulation, per Semafor's source. He then talked to Nvidia chief Jensen Huang about a set of principles the executives could carry into their next meeting with Trump. Zuckerberg circulated a draft before Tuesday's White House meeting, per Semafor; Huang rounded up support among the other guests. Trump then released the document himself.
Six companies signed: Meta, OpenAI, Anthropic, xAI, Google and Nvidia. Trump called the accord "morally binding." "And they understand that they have to self-police," he told reporters after the meeting.
The next day, the Federal Trade Commission confirmed to CNBC that it had opened an investigation into OpenAI, Anthropic and other AI companies over the potential dangers posed by their products. The spokesperson declined to name the others. OpenAI and Anthropic did not immediately respond to CNBC.
So the regulated handed Trump their own rulebook on Tuesday, and the regulator opened its file on Wednesday. What did the six actually promise, and who checks?
