WASHINGTON — As warnings mount within the artificial intelligence industry that AI could drive humanity toward catastrophe, U.S. President Donald Trump has dismissed federal regulation and signed an agreement on industry-led "self-regulation." The move is read as a decision to press the accelerator on development while adopting only the rules companies set for themselves, based on a view that the nation's fate hinges on the AI race with China.
After a lunch with the chief executives of major U.S. AI companies at the White House on the 29th, Trump said there is "a belief that there has to be a tremendous amount of self-regulation." He said regulation "is happening automatically through the Justice Department, the FBI and others, but self-regulation is very important." The remarks are taken to mean that voluntary controls by companies, on top of existing enforcement by the Justice Department and the FBI, can protect humanity from the threats posed by AI.
Bloomberg gave weight to the lunch, saying that if concerns about AI become reality, the meeting could rank as one of the most important in human history.
The lunch drew a full turnout of the chief executives of the U.S. AI companies that dominate the global industry. Attendees included Nvidia's Jensen Huang, Tesla's Elon Musk, Anthropic's Dario Amodei, Microsoft's Satya Nadella, Alphabet's Sundar Pichai and Meta's Mark Zuckerberg, along with OpenAI President Greg Brockman. Leaders from the defense and space sectors also took part, including Palantir Chief Executive Alex Karp and NASA Administrator Jared Isaacman. Twenty people attended from the private sector, while the administration was represented at the table by figures including Vice President J.D. Vance, House Speaker Mike Johnson (R-La.), Treasury Secretary Scott Bessent and Commerce Secretary Howard Lutnick.
Notably, Trump posted a seating chart for the participants on Truth Social before the event. According to the chart, Trump placed Huang, who has been unsympathetic to calls for slowing AI development, on his right. Musk sat on his left. Microsoft's Nadella and Amazon founder Jeff Bezos were also seated relatively close.
Amodei, who has argued for slowing down, was placed far away. He sat farther from Trump than Pichai, as well as Broadcom Chief Executive Hock Tan and Micron Chief Executive Sanjay Mehrotra. Anthropic, however, was the only company to send two people to the lunch, with Chief Computing Officer Tom Brown also attending, signaling a willingness to engage actively with the White House. Apple, which has been seen as lagging behind rivals such as Google and Meta in developing comparable AI models, was not even on the invitation list.
After the event, Trump released a document titled the "White House Superintelligence (SI) Accord." The paper, headed "Joint Commitments on Frontier Responsibility," set out the industry's self-regulation and carried the signatures of Trump, Pichai, Amodei, Zuckerberg, Musk and Brockman.
In detail, the agreement laid out four control and audit frameworks. It first specifies building "robust internal controls to monitor model capabilities and alignment in areas such as cyber and bio security and chemical threats, and to prevent hacking in unintended ways." It also states that companies "must form internal teams so that all controls and monitoring work as intended and problems can be resolved when they arise." In addition, it says companies "must work with independent external auditors and others to assess whether control and monitoring systems are functioning properly" and "must establish an independent committee under the board to ensure that all identified problems are corrected."
The Trump administration therefore appears set to leave AI risks to self-regulation for now while supporting development. Trump stressed that day that the United States is "ahead by a very big margin" in AI and "will maintain that gap." He also said AI companies would contribute to local communities, addressing growing resident opposition to data center construction.
Trump also signed an executive order changing "AI" to "SI." He explained that the term SI "more appropriately represents the possibilities, the potential and the rapidly advancing capabilities of the technology." Musk, who has returned to Trump's camp, provided a light moment by saying "AI" before hurriedly correcting himself to "SI."
