Huang told 12,000 people at Dreamforce the industry does not need new AI laws
Anthropic CEO Dario Amodei and Nvidia CEO Jensen Huang shared differing views on AI safety during Salesforce CEO Marc Benioff’s Dreamforce keynote in San Francisco on Tuesday, Ashley Capoot and Jordan Novet at CNBC report.
Amodei appeared first. Huang followed shortly after. The audience was about 12,000 at Moscone Center.
Dreamforce is Salesforce’s big annual customer conference. CNBC notes it draws about 50,000 people to downtown San Francisco.
A false choice claims you must pick only one of two good things. A frontier slowdown means labs agreeing to improve models more slowly.
Amodei took the stage days after his weekend essay urging the industry to slow the pace of model development. “I think that’s the way to lead the industry forward, to set an example, to say that everyone can always be better,” he told Benioff.
OpenAI’s Sam Altman, SpaceX’s Elon Musk, and Google DeepMind’s Demis Hassabis had voiced support. Amodei floated a three-step plan to temper how quickly model capabilities improve without “sacrificing commercial advantage or the United States’ lead in AI.”
Huang, CEO of “the world’s most valuable company,” said market forces already exist and that the industry doesn’t need new laws or regulations. Speed and safety, he said, are not mutually exclusive, though companies should pace themselves until they are sure they are releasing something the market will appreciate.
“It’s a false choice,” Huang told Benioff. “You could definitely have both at the same time.”
“Run as fast as you can,” he said. “But if you feel at any given point in time the company’s out of control or the product’s not going to be safe, take a pause and make sure you get it right.”
The pause he described is a company checking its own product, not an industry-wide brake and not a statute.
The same Tuesday, Salesforce and Nvidia announced a new reasoning model for Agentforce, Salesforce’s AI agent product line, built on Nvidia’s Nemotron, open-weight models anyone can download and run. Late last month Salesforce and Anthropic expanded their partnership under “Claudeforce,” a way for salespeople to reach data from inside Anthropic’s Claude chatbot.
Benioff has longstanding relationships with both CEOs. To reporters Tuesday: if labs feel they should slow down, slow down; if they feel they should speed up, speed up; “then they should be held accountable.”
Huang sells the processors that train and run the models. A broad training pause would leave product unsold. On this stage he put that commercial fact into policy language: no new laws, run as fast as you can, and pause only if your own company looks out of control.
The host’s rule is accountability after the fact. The chipmaker’s rule is no statute at all.