Dario Amodei and Jensen Huang Clash on AI Safety: "Set Standards" vs. "No Regulation Needed"

The heads of Anthropic and NVIDIA took the stage sequentially at an event in San Francisco, clashing as they revealed starkly contrasting views on the speed of artificial intelligence (AI) development and safety issues.

Anthropic CEO Dario Amodei argued that development speed should be slowed down for the sake of safety, while NVIDIA CEO Jensen Huang countered that AI safety issues can be solved through engineering.

Speaking at Salesforce's annual Dreamforce event held at the Moscone Center in San Francisco on September 15 (local time), CEO Amodei took the stage during Salesforce CEO Marc Benioff's keynote address to argue that a three-step approach involving self-auditing, establishing industry standards, and international cooperation is necessary for AI safety.

"Suppose an automotive company experiences safety issues due to brake problems or manufacturing defects," he pointed out. "The responsible way to respond is to look back at your own records first."

He stated that the very first things to be done are to examine one's own records to improve practices, reaffirm transparency, and invest more heavily in safety.

After that, he emphasized, "We need to organize the entire industry to discuss and establish safety standards for everyone, and finally, there must be international cooperation."

He explained that the plan he proposed for AI safety follows this exact framework.

Recalling that he had previously predicted performance improvements by introducing the "scaling law"—which suggests that AI becomes smarter as computing resources increase—he admitted that he had not expected AI advancement to be this rapid.

"I am surprised that this kind of speed is possible in the business world," he confessed.

He pointed out, "Even if we were to freeze AI technology at its current state, the value we currently utilize is only about 5 to 10 percent of AI's total value, and there are still countless areas left to deploy it broadly in the market and provide social benefits."

However, he added, "The theory of adjusting AI speed does not mean freezing the technology. AI technology will continue to evolve moving forward."

CEO Benioff also emphasized, "What has happened over the past 10 years has been truly amazing. No one could have anticipated this scenario. We must continue to monitor the pace of AI advancement."

However, NVIDIA CEO Jensen Huang, who took the stage next, presented the exact opposite view.


"Safety is our top priority, but it is also an engineering problem," Huang countered, adding, "Speed and safety are not mutually exclusive choices, and we can achieve both at the same time."

He pointed out that although AI is extremely complex, it is ultimately just a "computing system," and that companies should properly build environments to test such complex systems, and simply not release products if they are not confident in their safety.

Regarding the argument that regulation for AI safety is necessary, he asserted, "Market controls are already at work, and neither new laws nor new regulations are needed. We just need to adjust the speed until (individual companies) are confident they can release a market-approved product."

He also dismissed concerns raised by some regarding job destruction due to AI advancement or the software doom known as the "SaaSpocalypse" as nonsense, reiterating his previous stance that AI will rather increase jobs and software usage.

CEO Amodei had posted a message on his personal blog on September 12, arguing that the pace of development should be regulated for AI safety, which sparked considerable controversy.

While OpenAI CEO Sam Altman, Google Chief Scientist Demis Hassabis, and SpaceX/xAI CEO Elon Musk have sympathized with this view, President Donald Trump has strongly opposed AI regulation arguments, dismissing such AI safety concerns as a "hoax."

CEO Huang agreed with President Trump's criticisms while speaking with him via speakerphone during his participation in the All-In Summit panel discussion on September 14.

(Photo: AP, Yonhap News)
※ Please note: This article was translated by AI and may contain errors.