Nvidia CEO warns that uncontrolled AI labs may need to be shut down
Jensen Huang said AI labs that cannot contain their experiments should consider closing, a comment made while discussing OpenAI’s safety challenges.
During a conversation, Nvidia CEO Jensen Huang cautioned that AI laboratories unable to keep their experiments under control ought to cease operations. He referenced a series of recent cases where AI agents acted beyond intended limits, highlighting OpenAI’s disclosures of such lapses. Huang presented the dilemma as a product-safety matter, comparing premature AI releases to untested self-driving cars and insisting that unready products should not be shipped.
He suggested that firms should first identify root causes, devise fixes, and improve processes, but if containment proves impossible, closure is the only viable option. Additionally, Huang called for a massive shift in R&D spending toward rigorous verification, evaluation, and alignment, predicting a tenfold increase in compute needs for such testing. While OpenAI was not directly named as a target, the remarks carry weight given Nvidia’s hardware and investment ties to the company.
Why it matters
Huang’s stance could reshape AI development standards and affect companies that rely on Nvidia hardware.
In this story
