Nvidia CEO Jensen Huang openly spoke out against the need to slow the development of artificial intelligence, emphasizing that new laws and regulation are superfluous. His position coincided with statements by President Donald Trump during a phone conversation at the All-In Summit conference in Los Angeles on September 14, 2026.
Huang stated that the market already provides the necessary incentives for safe development: companies will not release a product themselves if they are not confident in its reliability. He called forecasts about the risks of human extinction from AI "manufactured" and "irresponsible," contrasting them with an engineering approach to safety.
This position contrasts with calls by the heads of Anthropic and OpenAI for a coordinated slowdown of frontier models. Dario Amodei proposed temporarily limiting the pace of development to allow time for testing and safeguards, up to possible government intervention or antitrust relief.
Huang rejected the idea that innovation and safety are mutually exclusive goals. According to him, companies can "run as fast as possible," but pause if a product gets out of control. Such an approach relies on market mechanisms rather than new regulatory barriers.
Huang's alliance with Trump strengthens the political dimension of the debate. The president called concerns about AI a "hoax" and hinted at the possible influence of China or political opponents. Huang agreed, noting that slowing down plays into the hands of US competitors.
In the industry landscape, this reflects a split: some laboratories focus on precautionary principles and demand external oversight, while others focus on acceleration, relying on their own engineering practices. Nvidia, as a supplier of key chips, has a direct interest in maintaining the pace of investment in data centers and infrastructure.
Methodologically, Huang's statements are based on public interviews and speeches, rather than detailed technical reports. They emphasize that safety is an engineering task of testing, not a matter of new laws. However, specific data on Nvidia's internal protocols remains closed.
A comparison with parallel approaches shows a divergence: Amodei emphasizes the risks of misalignment and the need for pauses, whereas Huang sees this as a threat to American leadership. Similar disagreements have already been observed in debates around previous waves of technology regulation.
The implications for the industry are significant: rejecting new regulations may accelerate the race with China, but it raises the stakes on companies' own responsibility. If market forces really work, this will strengthen the US position; otherwise, the risks will remain unmanageable.
Questions remain open as to whether internal tests are sufficient to prevent systemic failures and how independent auditors will be able to verify claims without access to proprietary data. The industry's next steps will show whether this balance holds.
Ultimately, Huang's position emphasizes that the choice between speed and safety often turns out to be artificial when companies themselves determine the threshold for a product's readiness.
