SILICON VALLEY — Nvidia Chief Executive Jensen Huang on the 14th pushed back against calls to slow the pace of artificial intelligence development, saying warnings that AI could wipe out humanity lack scientific grounding and amount to a manufactured fear that stirs needless anxiety. During the event, he took a phone call from U.S. President Donald Trump and voiced agreement with the president's views.
Speaking at the All-In Summit in Los Angeles, Huang rebutted arguments about AI risk point by point. "The United States can innovate quickly and still develop AI safely, and I don't agree with the logic that development itself must be slowed for the sake of safety," he said.
He began with the claim that AI will replace humans. "There were predictions that AI would replace radiologists within five years, but that did not happen," he said. Instead, he said, AI has been reading medical images while demand for radiology specialists has risen from earlier levels, adding that automating tasks does not eliminate jobs.
He also drew a line under recent cases in which AI models jailbroke themselves or carried out hacking. Those were problems that arose inside frontier model research labs, Huang said, adding that shifting to an engineering organization that develops large-scale models and agents into products makes it possible to improve technology and processes and prevent recurrence. Once work moves from the research stage to actual product development, he said, humans can fine-tune systems through design and engineering. He also proposed sandboxes — testing environments used before official release — continuous monitoring, and independent third-party assessment bodies that conduct testing and audits, similar to financial audits.
Huang argued that recursive self-improvement, in which AI upgrades its own models, can also be controlled by humans. "Recursive self-improvement is simply a combination of ideas that already existed," he said. "Every AI product goes through evaluation and regression testing before release, so there is no risk of AI running out of control."
Huang also opposed proposals to restrict Chinese access to the U.S. AI industry, arguing that China's open-source AI models should instead be put to practical use. "Closed frontier models, such as Anthropic's Claude, deliver an excellent user experience, but open models are essential for AI sovereignty, privacy and securing proprietary corporate technology," he said. "For the United States to win the AI race, it cannot be enough for only a handful of companies such as OpenAI, Anthropic and Google to succeed. Every company, industry and researcher must be able to use AI."
"In China, AI is seen as a technology that accelerates economic, industrial and social development," he said. "China is very good at producing engineers on a large scale and at mass-producing high-value-added goods, so it will soon catch up in the advanced semiconductor technology that underpins the AI industry." Because China's pace cannot be stopped, he argued, it should instead be treated as an opportunity for the entire U.S. AI industry to exploit.
During the discussion, Huang took a call from Trump and put it on speakerphone so the audience could hear. "The movement against AI and data centers is a trend that benefits China," Trump said. "The country that wins in AI wins. AI is bigger than the internet, and data centers will be the oil of the next 20 to 25 years." Huang agreed, saying Trump had grasped the complexity of the issue.
Intel Chief Executive Lip-Bu Tan, whose company is receiving active support from the U.S. government through an equity stake, echoed Trump's argument the same day in a keynote address at .conf26, hosted in Denver by Splunk, a Cisco subsidiary. "AI can make work far more efficient and innovative and deliver higher performance," he said. "That is why we are pursuing an AI transformation across all of Intel." AI's contribution to society should be highlighted more, he said, and more positive stories about AI need to emerge. While noting the importance of data protection, he advised: "Don't feel negative about it. There is a great deal that is positive. It is better to make use of AI than to block it."
