TL;DR — Key Takeaways
- President Trump rejected calls to slow frontier AI development during an unexpected phone call with Nvidia CEO Jensen Huang at the All-In Summit.
- Huang said concerned AI labs could voluntarily pace their own development, while Anthropic CEO Dario Amodei is seeking independent evaluations and coordinated safety standards for frontier models.
- The dispute has become an open industry split over how to manage increasingly capable AI.
President Donald Trump rejected growing calls to slow frontier AI development Monday during an unexpected phone call with Nvidia CEO Jensen Huang, calling fears that AI could take over the world a “hoax.”
Huang was discussing AI safety onstage at the All-In Summit in Los Angeles when Trump called. Huang put the president on speakerphone, turning the nearly five-minute conversation into a public response to the ongoing safety debate that had intensified over the weekend.
“The robots are not going to be taking over the world,” Trump said. He said slowing U.S. AI development could benefit China, while acknowledging that the technology should be handled carefully without stopping the industry. Huang agreed that the U.S. should not allow an AI slowdown.
Before Trump called, Huang had already pushed back on proposals to deliberately pace AI development earlier in his interview at the summit. He said labs concerned that their systems were getting out of control could voluntarily pause or pace their own work. At the same time, Huang said AI safety remains important and distinguished legitimate concerns from predictions about catastrophic future outcomes that he said lacked sufficient evidence.
Huang’s remarks at the All-In Summit came after Anthropic CEO Dario Amodei published an essay calling for companies to slow the rate at which they increase frontier model capabilities so that safety research, testing and oversight have time to keep up.
Amodei stressed that “pacing” does not mean stopping model training or technical progress. His proposal starts with frontier labs giving independent evaluators ongoing access comparable to employees so they can verify safety practices, examine training processes and report incidents. Anthropic said it will adopt that measure itself and wants governments to require other frontier developers to do the same.
The second step would involve frontier AI companies in democratic countries coordinating on common safety standards and limits on unchecked capability advances, with government involvement where necessary. Amodei ultimately calls for attempts at international coordination as well. He said that even one or two additional years before models reach critical levels of capability could give researchers substantially more time to improve alignment, interpretability and model evaluations.
Trump reinforced his opposition to an AI slowdown in a series of Truth Social posts Monday, arguing that the U.S. already has sufficient authority to address wrongdoing involving AI and that additional restrictions could undermine the country’s position relative to China. He also said the only AI “guardrails” needed were a “strong and smart” president and said there was a “SICK conspiracy” against AI and data centers.
The dispute has become an open industry split, with major AI companies taking different approaches to how increasingly capable models should be developed and controlled. On Monday, Microsoft published a draft code of conduct for its MAI models, including safeguards to keep the models under human control. Unlike Amodei’s proposal, however, Microsoft’s framework governs how its own models are trained and deployed rather than calling for an industry-wide slowdown in development.
