Anthropic, Nvidia CEOs Split on AI Slowdown at Dreamforce
Anthropic chief executive Dario Amodei and Nvidia chief executive Jensen Huang offered directly opposing prescriptions for frontier AI development on Tuesday at Salesforce's Dreamforce convention in San Francisco. The Guardian
Amodei reiterated his call for a slowdown of AI development on stage at Dreamforce. He called for embedding third-party evaluators inside AI companies. He called for coordinating AI safety standards among democratic countries. He called for eventual larger global coordination on AI safety. He said Anthropic has committed to independent evaluators.
The Dreamforce remarks followed a 3,800-word essay in which Amodei called for a global slowdown of artificial intelligence development. The New York Times Amodei has called for AI labs to slow down frontier AI development due to safety concerns. He is not alone in that appeal. Amodei, xAI's Elon Musk and OpenAI's Sam Altman have all recently called for a slowdown in AI development.
That debate has acquired sharper edges inside the labs. Former Anthropic researcher Jacob Coxon resigned with a public warning that AI could lead to extinction of humanity within a decade. Altman plans to delay OpenAI's IPO until next year over safety concerns. Both OpenAI and Anthropic are headed for initial public offerings.
Huang argued against slowing AI development shortly after Amodei at Dreamforce, saying "Run as fast as you can." Speed was the point. He said new AI regulations aren't necessary. He advised AI companies to wait to release products until they know they are safe rather than seek US government intervention.
Huang also rejected economic alarm about automation. He called the idea that AI is going to destroy jobs "completely nonsense". He said he envisions a future where every company and enterprise would become an AI company.
The exchange reached beyond the two executives. At the All-In summit in Los Angeles one day before Dreamforce, Donald Trump called Huang on stage and called growing concern about AI a "hoax". Trump asserted on the call with Huang that an AI slowdown would only benefit China.
Dreamforce runs September 15-17 in San Francisco and on Salesforce+. Salesforce The Main Keynote was scheduled 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. Dreamforce also includes a session titled 'A Conversation with Marc Benioff and Sam Altman' about OpenAI advancing the AI frontier and partnering with Salesforce.
The broader context here is a collision between two theories of safety. One holds that external scrutiny and coordinated restraint among capable labs can reduce risk before systems are deployed. The other holds that internal testing by developers, without new state rules, is sufficient and faster. In my view, the Dreamforce sequence matters because both positions were stated in the same venue, hours apart, to an enterprise audience that will buy and deploy the resulting systems.
Looking at what this means for diplomacy, the fault lines are already visible. Amodei's formula moves from coordination among democratic countries toward eventual larger global coordination. That sequencing will be difficult to negotiate. It raises questions about who sets standards, who gets access to evaluators, and how competition with China shapes willingness to pause. Trump's intervention frames any slowdown as a relative gain for Beijing. For practitioners, the near-term signal is fragmentation: voluntary evaluator commitments from some labs, resistance to formal regulation from hardware leadership, and IPO calendars proceeding under public safety warnings.
