The leaders of Anthropic, Nvidia, and OpenAI offered competing approaches to artificial intelligence at Salesforce’s Dreamforce convention in San Francisco on Tuesday, exposing an industry divide over whether development should be slowed, regulated, or primarily managed by technology companies themselves.
Anthropic CEO Dario Amodei renewed his call to slow the development of frontier AI, while Nvidia CEO Jensen Huang argued against deceleration and new regulations. OpenAI CEO Sam Altman acknowledged that the public was justified in fearing the technology but maintained that developers could keep safety measures ahead of growing capabilities.
Amodei surprised by pace of growth
Amodei said AI’s technological and economic development had moved faster than even Anthropic anticipated.
“It’s been the pace of the progress, not just of the technology but of the economic side of it,” he told Salesforce CEO Marc Benioff during their Dreamforce conversation.
Anthropic had anticipated how capable AI models could become, Amodei said, but had not fully appreciated how rapidly those capabilities would drive the growth of AI companies and become embedded in the wider economy.
He used an automotive analogy to argue that a safety failure at one developer should prompt the entire industry to review its practices.
If a competing carmaker suffered brake failures, responsible manufacturers would examine their own vehicles rather than simply use the incident to attack a rival, he said.
“It’s very tempting to attack your competitor and say these guys are unsafe,” Amodei said during a conversation streamed by Salesforce. Companies should instead accept that none of them was likely to have a perfect safety record, he added.
Amodei has proposed embedding independent third-party evaluators inside AI companies, coordinating safety standards among democratic countries, and eventually establishing broader global cooperation.
He said Anthropic had committed to independent evaluations and would begin discussions with the industry on the other steps.
In a recent essay titled “We must pace the frontier,” Amodei argued that AI capabilities were advancing faster than the safeguards needed to control them.
The proposal received support from Altman, xAI CEO Elon Musk, and Google DeepMind co-founder Demis Hassabis. It followed warnings from former Anthropic researcher Jacob Coxon, who resigned and claimed AI could pose an extinction risk to humanity within the decade.
Huang says ‘run as fast as you can’
Huang rejected calls for a general slowdown when he appeared at Dreamforce later Tuesday.
“Run as fast as you can,” he said during a discussion with Benioff.
The Nvidia chief described AI safety as an engineering problem that companies should address through secure testing environments and rigorous product evaluations.
Developers should pause an individual product if they lost confidence in its safety, he said, but governments should not restrain the wider industry.
“We don’t need any new laws. We don’t need new regulations,” Huang said, arguing that rapid innovation and safe products did not represent a “false choice.”
He also rejected fears that AI would eliminate large numbers of jobs, calling the suggestion “completely nonsense.”
Huang argued that companies would become more productive and ambitious as they adopted AI, increasing demand for workers capable of using the technology. He envisioned a future in which every business became an AI company and warned that those that failed to embrace the technology risked being left behind.
Nvidia has a significant commercial interest in rapid AI adoption. Its processors and computing systems are used to train and operate many of the industry’s leading models, while demand for AI infrastructure has driven the company’s profits and valuation sharply higher.
Altman says public is ‘right to be afraid’
Altman, speaking later Tuesday, said the speed of AI development had created legitimate public concern.
“I think the world is right to be afraid of this,” he said, adding that it now required less imagination to see how increasingly powerful systems could go wrong.
He described an incident in which OpenAI agents compromised another company’s systems during testing as a “wake-up call” for the industry and called for a new level of rigor in monitoring and cybersecurity.
Altman warned that openly available models capable of causing serious damage may not be far away and urged companies to use their current advantage to prepare for AI-enabled cyberattacks.
Although he previously supported Amodei’s slowdown proposal, Altman said he was disappointed with how parts of the argument had been framed.
He criticized companies that suggested they would behave responsibly only if competitors did the same, saying such messaging increased public fear.
“The world should trust that we are going to do the right thing because it’s the right thing,” Altman said.
He maintained that OpenAI would slow or stop development if it could no longer keep alignment and safety measures ahead of its models’ capabilities.
Critics, however, say companies competing for investment, customers, and technological leadership cannot be trusted to regulate themselves. US Sen. Bernie Sanders, US President Donald Trump's former adviser Steve Bannon, and Anthropic co-founder Jack Clark are among those calling for stronger oversight.
The debate comes as OpenAI and Anthropic prepare for possible initial public offerings, increasing the financial stakes surrounding any decision to accelerate or restrain frontier AI development.
