SILICON VALLEY — Anthropic Chief Executive Dario Amodei and Nvidia CEO Jensen Huang delivered opposing messages on artificial intelligence regulation from the same stage. Amodei stressed the need for globally agreed standards, while Huang countered that the matter can be left to the market and that no new rules are needed. The exchange extended a running dispute between AI model developers and AI chipmakers over regulation.
Salesforce CEO Marc Benioff spoke with Amodei during a keynote session at the Dreamforce conference in San Francisco on the 15th. Benioff raised an essay Amodei published on the 12th arguing for a measured pace in AI development and for greater discipline in the field. At the event, Benioff unveiled AIforce, which brings all of the company's capabilities into the workplace, and discussed the future of AI with Amodei and other chief executives.
Asked how he views the current pace of AI, Amodei said an analogy would be easier to follow and pointed to the auto industry. "Suppose a car company has a safety incident because of a brake problem or a manufacturing problem — how would you respond?" he said. "It can be tempting to attack a competitor by saying that company isn't safe, but the more responsible response is to look at your own record first. To check whether your own company has had issues too."
Amodei added: "Even if there was no incident, it's certain that we aren't perfect. Let's improve our practices. Let's commit to transparency. Let's invest more in safety." He proposed: "And then let's discuss what we can do to set standards for the industry as a whole. Let's think about what we can do to raise the bar for everyone." He said an international element also has to be included, describing that as the core of the three-stage plan he laid out in his essay. "Stage one is what we have already committed to, and we plan to talk with the industry about the rest," he said. "I think this is how you advance an industry. Setting an example that everyone can always do better."
Asked what had surprised him most over the past decade, Amodei cited the pace of progress. "A few of my co-founders and I formalized the scaling laws, which show that AI models get steadily smarter the more computing resources you put into them," he said. "We were able to predict roughly how capable the models would be, but we didn't expect that companies would grow this fast or become central to how the world works. That part is genuinely surprising." His argument is that standard-setting and oversight work are needed to manage AI advancing faster than anticipated.
Huang, who appeared after Amodei left, delivered the opposite message. Asked how he viewed the debate over slowing AI development and safety concerns that had emerged over the past week, Huang said: "Safety is the most important thing in many respects, but safety is an engineering problem." He said the principle is clear that a product should not be released if its safety has not been established, and that companies should moderate their pace until they can ship something the market finds satisfactory. "But market forces already exist. New laws or regulations aren't needed," he said. "All that's needed is deciding when a company runs as hard as it can."
Huang's remarks emphasized that securing AI performance and safety should be left to individual companies rather than addressed through uniform regulation. "You can have both" innovation and safety, he said. "Run as fast as you can, but if at some point you feel things are out of control or the product doesn't seem safe, pause and fix it properly."
Huang also dismissed talk of an AI-driven software crisis or jobs crisis as nonsense. "Many people were surprised, but I wasn't. I was the first CEO outside the software industry to say the software apocalypse was nonsense," he said. "A layer is being added on top of software, and that will make software far more usable. The idea that AI will destroy jobs makes no sense. AI is becoming more productive and more useful."
Huang stressed that companies must actively adopt AI to avoid falling behind. "You represent your companies and your countries. You don't want to fall behind," he said. "This is far too important a technological revolution. It's a technology with tremendous capability, so you have to engage with it and learn it. Even if it doesn't work right away, don't give up. There's a huge industry working behind it to improve this technology constantly. Keep trying." He added: "Companies will become more ambitious than ever. With AI-driven productivity gains, there are no limits for us. There are no limits for any industry or any country. So engage with AI. Don't fall behind."
