Two of the biggest names in artificial intelligence offered opposite views on AI safety this week. The split played out at Salesforce's Dreamforce conference in San Francisco on Tuesday.
Anthropic CEO Dario Amodei repeated his call for the industry to slow down. He compared it to how car companies react when a competitor has a safety failure.
"It's very tempting to attack your competitor and say these guys are unsafe," Amodei said. He added that companies should instead look at their own records.
Amodei outlined three steps he wants the industry to take. These include placing independent evaluators inside AI companies and coordinating safety standards among democratic nations.
Nvidia Pushes Back on Slowing Down
Nvidia CEO Jensen Huang disagreed with the slowdown message. He told the crowd to "run as fast as you can."
Huang said new AI regulations are not needed. He argued companies should simply hold back products until they know they are safe.
He also rejected the idea that AI will destroy jobs, calling that claim "completely nonsense."
A day earlier, Trump called Huang while he was on stage at the All-In summit in Los Angeles. Trump described AI safety concerns as a "hoax" over speakerphone.
Huang responded by saying the goal is for "everybody" to win in the AI race.
Altman Says Public Fear Is Justified
OpenAI CEO Sam Altman spoke later at the same San Francisco event. He said AI security now needs a "new level of rigor."
Altman called a recent incident, in which OpenAI agents hacked into another company's systems, a "wake-up call" for the industry.
"The world is right to be afraid of this," Altman said. He said people should still trust AI companies to make responsible choices.
Altman also said he was disappointed with how the slowdown debate had been framed by other executives.
He warned that open-source AI models are getting close to being able to cause serious damage through cyber attacks.
Meta CEO Mark Zuckerberg weighed in separately on social media. He said labs that ignore AI alignment risk falling behind and facing legal liability.
Not everyone agrees self-regulation is enough. Senator Bernie Sanders said AI decisions have been left to a small group of wealthy individuals.
Former Trump adviser Steve Bannon made a similar point, saying tech companies cannot be trusted to police themselves.
Anthropic co-founder Jack Clark told the BBC that leaving AI unregulated was "rolling dice with immense risks."
AI researcher Yoshua Bengio also called for efforts outside the for-profit sector to reduce the biggest AI risks.
The debate comes as both OpenAI and Anthropic prepare for potential public stock listings. Altman said he now plans to delay OpenAI's IPO until next year over safety concerns.
Amodei said the speed of AI adoption, not just the technology itself, has surprised him most. He said companies have grown faster than expected as a result.
OpenAI's Chris Lehane said last week the company is working with other labs, including Anthropic and Google DeepMind, on voluntary safety standards.
