AI system safety should be viewed as an engineering challenge, with decisions on development pace and potential pauses left to the companies themselves. This statement was made by Nvidia CEO Jensen Huang during a conversation with Salesforce CEO Marc Benioff at the Dreamforce conference.
Benioff asked the Nvidia chief about the ongoing industry debate over slowing AI development.
“Safety is paramount. But safety is an engineering problem,” Huang replied.
According to him, developers create software and computing systems, so they must first establish reliable environments for testing them. If a company is unsure about a product’s functionality or safety, it should not be released.
“Market forces already exist. We don’t need new laws. We don’t need new regulations,” he stated.
The Nvidia CEO also dismissed the notion of a false choice between innovation speed and safety. In his view, companies can develop technologies rapidly but should pause if they lose control over the process or doubt the product’s safety.
Huang’s remarks came days after Anthropic CEO Dario Amodei called for slowing the development of the most powerful AI models. His position was supported by OpenAI CEO Sam Altman and Elon Musk.
On September 14, amid the debate, shares of several AI-related companies and semiconductor manufacturers fell. Market participants assessed the potential impact of a slowdown on demand for computing infrastructure.
The issue of coordination among developers is complicated by antitrust laws. On September 11, OpenAI asked members of the U.S. Congress to clarify whether joint efforts to slow the development of advanced models would be considered anti-competitive.
Amodei himself described regulating all American developers of advanced models as the most effective option. For voluntary coordination among competitors, he suggested government involvement and a narrow exemption from antitrust restrictions.
In July, Nvidia, Meta, Microsoft, and 22 other companies urged U.S. authorities not to impose general restrictions on open-weight AI models. Huang then stated that such systems promote innovation, cybersecurity, and the dissemination of technology.
