Anthropic Co-Founder and CEO Dario Amodei and Salesforce CEO and Co-Founder Marc Benioff attend the Dreamforce 2026 Summit in San Francisco, California, USA on September 15, 2026.
Carlos Barria | Reuters
Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman offer a different take on the safety of artificial intelligence. Nvidia A conversation with CEO Jensen Huang sales force Dreamforce conference just days after the discussion went mainstream.
Amodei was first on stage with Salesforce CEO Marc Benioff for his annual keynote address. The keynote appearance followed an essay Amodei released over the weekend in which he implored the industry to slow down the pace of model development.
“I think that’s a way to move the industry forward, to lead by example, and to let everyone know that they can always be better,” Amodei told Benioff before a crowd of about 12,000 people at San Francisco’s Moscone Center on Tuesday.
Industry executives such as OpenAI CEO Sam Altman and SpaceX CEO Elon Musk; google DeepMind Chairman Demis Hassabis quickly voiced support for Amodei’s proposal after industry researchers last week sounded the alarm about the potential for AI to cause catastrophic harm. Amodei proposed a three-step plan to slow the rate of improvement in model capabilities without “sacrificing commercial advantage or America’s lead in AI.”
Fans who showed up with Benioff shortly after Amodei have a different perspective. The CEO of the world’s most valuable company said market forces already exist and the industry doesn’t need new laws or regulations. Huang added that speed and security are not mutually exclusive, but companies should pick up the pace until they are confident they are releasing something the market appreciates.
“That’s a false choice,” Huang told Benioff. “You can definitely have both at the same time.”
Huang said it’s up to model developers to act responsibly with their products.
“Run as fast as you can,” he said. “But if at some point you feel like the company is getting out of control or the product is no longer safe, pause for a moment and see if you can do it right.”
The day before, Hwang took to the stage at the All In Summit in Los Angeles and received a phone call from President Donald Trump. During his speech, the president expressed concern about people trying to obstruct the construction of AI data centers.
“We’re not going to allow that to happen,” Huang told Trump.
Dreamforce has been one of the software industry’s premier events for more than a decade, drawing approximately 50,000 people to downtown San Francisco and occupying a large swath of the local area. Benioff’s keynotes always include multiple prominent guests, often people at the center of the technology industry at that particular time.
Benioff has had long-standing relationships with both Mr. Huang and Mr. Amodei.
On Tuesday, Salesforce and Nvidia announced a new inference model for the software company’s Agentforce platform built on one of Nvidia’s open-weight Nemotron models. Late last month, Salesforce and Anthropic expanded their strategic partnership and announced Claudeforce, a new way for sales reps to access critical data directly from within Anthropic’s Claude chatbot.
“If you feel like you should slow down, you should slow down. If you feel like you should speed up, you should speed up. And then you should be accountable for that,” Benioff told reporters Tuesday about how model companies should proceed.
Mr. Altman and Mr. Benioff’s appearances took place later in the day. During a fireside conversation between the two CEOs, Altman said that given the capabilities of existing models, it was imperative to prioritize safety and oversight over functionality.
“Some companies say things like, ‘We’re only responsible if other companies are responsible,'” Altman said. People then ask what the company is going to do to stay safe and accountable “at all costs,” he said.
“There shouldn’t be any restrictions attached to it,” Altman said.
WATCH: Sen. Mark Warner: We need to use AI leaders to push for regulation
