A product conference at software giant Salesforce turned into the battleground on the new debate on whether there should be a slowdown in artificial intelligence.
The Dreamforce 2026 summit in San Francisco featured leading executives including Anthropic CEO Dario Amodei, OpenAI chief Sam Altman and Nvidia CEO Jensen Huang, as Meta Platforms CEO Mark Zuckerberg weighed in on social media.
Under the gentle questioning of Salesforce CEO Marc Benioff, Amodei made an analogy. “Let’s say you’re running a car company. And another car company, not yours, has some kind of safety incident. Something goes wrong with the brakes, something with the manufacturing… Now, you believe you have the best safety record in the industry. What do you do, how do you approach it?”
Amodei said the three-step plan he has is first to improve processes at his own company, then to set industry standards and then have an international component to it.
Altman, in a separate Q&A with Benioff, said the why-now is that in some ways the models are now smarter than people. “When people say, ‘my probability of doom is x or y or whatever, it hits in a different way,” said Altman.
“People also feel the speed with which things are moving,” he said. The two worries, he added, are both a loss-of-control type accident as well as a concentration in power.
Huang, in his trademark leather jacket, said safety is paramount but an engineering problem. “If you build a product or a service and you’re not confident in its functionality, capability or safety, then don’t release it,” said Huang. “You pace yourself until you are confident you’re releasing something the market would appreciate.”
“We don’t need any new laws. We don’t need new regulations,” he added. “Run as fast as you can, but if you feel at any given point in time, the company’s out of control, or the products not going to be safe, take a pause and make sure you get it right.”
On X, Zuckerberg largely agreed with Huang. “Labs face significant liability if their models cause harm, so they have a strong incentive to prevent this as well,” wrote the Meta chief. “Meta delayed shipping [AI assistant] Muse for several months to focus on safety and security. We didn’t call for everyone else to do this before we would. We just did it as part of our day-to-day work because it was clearly the right thing for people and for us. I’m proud of the security foundations we’ve built.”
As for Salesforce, it rolled out a reasoning model called Koa based on Nvidia’s open-source Nemotron model and trained on synthetic data.
The most popular semiconductor ETF on Tuesday rose 0.3% after a 5.6% slide on Monday, while the software ETF had the reverse move, falling 1% on Tuesday after a 5% gain on Monday. In early Wednesday premarket trade, the chip ETF rose another 1% while the software ETF retreated by 1%.
JPMorgan’s Asia tech analyst Gokul Hariharan weighed in on the debate. “While AI safety concerns are a near-term hiccup, we believe that frontier AI model training is unlikely to slow down, and the fast pace of AI evolution with [recursive self-improvement] etc should create new revenue [total addressable markets] beyond coding,” he said.
Comments