Should AI Slow Down? Two Tech CEOs Disagree
Anthropic boss Dario Amodei and Nvidia boss Jensen Huang gave opposite advice on Tuesday at Salesforce's Dreamforce meeting in San Francisco about how fast to build powerful AI. The Guardian
By powerful AI, they mean the newest and most capable systems. Amodei repeated his call to slow down. He said AI companies should let outside safety checkers work inside their labs, like building inspectors who check a house before people move in. He said democratic countries should agree on safety rules first, then build wider global rules later. He said Anthropic has already agreed to use independent checkers.
His remarks followed a 3,800-word essay calling for a world-wide slowdown. The New York Times Amodei has said labs should slow down because of safety risks. He is not alone. Elon Musk of xAI and Sam Altman of OpenAI have also recently called for a slowdown.
The debate has reached inside the labs. Former Anthropic researcher Jacob Coxon quit and warned in public that AI could cause human extinction within ten years. Altman plans to delay OpenAI's IPO until next year over safety concerns. An IPO is a first sale of stock to the public. Both OpenAI and Anthropic plan to hold IPOs.
Huang spoke shortly after Amodei and disagreed. He said "Run as fast as you can." He said no new AI rules are needed. He said companies should hold back a product until they know it is safe, rather than ask the U.S. government to step in.
He also rejected worry about jobs. He called the idea that AI will destroy jobs "completely nonsense." He said every company will in time become an AI company.
The dispute goes beyond the two men. At the All-In summit in Los Angeles one day earlier, Donald Trump brought Huang on stage and called rising concern about AI a "hoax." Trump said a slowdown would only help China.
Dreamforce runs September 15-17 in San Francisco and on Salesforce+. Salesforce The Main Keynote was set for Tuesday, September 15 at 10:00AM PT. It features Marc Benioff, Chair and CEO of Salesforce, with leaders from NVIDIA, Anthropic, Siemens, and more. The program also lists 'A Conversation with Marc Benioff and Sam Altman' about OpenAI pushing the AI frontier and partnering with Salesforce.
The broader context here is a clash between two safety ideas. One says outside checks and joint restraint by leading labs can cut risk before launch. The other says careful testing inside companies, without new government rules, is enough and faster. In my view, the Dreamforce moment matters because both views were stated in the same place, hours apart, to business buyers who will use these systems.
Looking at what this means for diplomacy, the splits are already clear. Amodei wants democracies to agree first, then bring in the wider world. That order raises hard questions about who writes the rules, who gets the checkers, and how rivalry with China affects any pause. Trump casts any slowdown as a win for Beijing. For now, the result looks split: voluntary checker pledges from some labs, pushback on formal rules from chip leadership, and IPO plans moving ahead alongside public safety warnings.
