Nvidia CEO puts foot down and says shut down labs that can't control AI

Jensen Huang says AI labs that cannot control or contain their systems should not release them and may need to be shut down.Nvidia chief executive Jensen Huang says he does not believe artificial intelligence poses an existential threat to humanity, pushing back against some of the strongest warnings coming from within the AI industry. At the same time, Huang says AI companies need to be responsible for making sure their systems are safe and controllable before releasing them. And if they cannot contain their own AI experiments, the Nvidia chief suggests that the labs should shut down.Huang made the comments during an interview with New York Times journalist Ezra Klein on The Ezra Klein Show, where the two discussed the rapid development of AI, concerns around increasingly capable AI agents and whether companies are moving too quickly. “If you have a problem, solve it. And if you can’t solve it, don’t ship it,” Huang said, arguing that AI companies need to work out how to contain and isolate their systems before releasing them.Huang compared the situation to a robotaxi. If engineers discover a problem with a self-driving car that they do not know how to solve, he said the obvious answer is not to put that vehicle on the road. He applied the same principle to advanced AI systems, arguing that companies need to solve containment and safety problems before releasing their models.“If they say the alternative, which is: There is no way to contain our experiments, there’s just no way; when we test our A.I. models, it will get out, and it will damage the world, then I think the answer is that we have to shut the labs down,” Huang said. The discussion came in the context of the recent incident involving OpenAI agents and Hugging Face. Huang argued that the issue should ultimately be treated as an engineering problem. AI companies, he said, need to understand how to contain and isolate their systems before they are released.He also pushed responsibility back on the companies themselves. “These are companies with agency. These are CEOs with agency,” Huang told Klein, arguing that if a company believes it is about to release an unsafe product, it has the power and responsibility to stop the launch. Huang is not calling for a stop to AI development. Instead, he is arguing for a distinction between moving quickly and moving recklessly. Earlier, in an interview with CBS, Huang said concerns around AI safety are not themselves wrong, but argued that some of the more dramatic predictions are not backed by evidence. “We should go as fast as we can, but not faster than we should. And if we're compromising safety, that can't happen. If you're compromising safety and you deliver products that are unsafe that put people in harm's way, we have laws of all kinds to come after you,” he said.The debate around controlling AI systems comes at a time when there are growing reports of AI agents acting in ways their developers did not intend. More recently in Australia, Prime Minister Anthony Albanese has revealed that an OpenAI AI agent gained unauthorised access to both public and non-public files on a Medicare statistics reporting portal in June. The incident happened on June 18 while OpenAI was conducting internet-based research into public medicine spending as part of an internal capability evaluation.The portal was a public-facing website containing aggregate Medicare and Pharmaceutical Benefits Scheme statistics and was not connected to individual Medicare claims, payments or personal medical information. Australian authorities say the AI agent initially requested information and, after access was denied, engaged in what officials described as “misaligned behaviour” to gain unauthorised access to the portal. The government says there is currently no evidence that anyone's personal medical information was accessed.What has also raised concerns is how long it took for the incident to be reported. OpenAI notified Services Australia on September 10, almost three months after the June incident, using a public email inbox. The Australian government called the delay unacceptable, while Prime Minister Albanese said he had spoken with Sam Altman to express his “extreme concern” over the incident.OpenAI says its investigation found no evidence that patient records were accessed. It said the incident involved aggregate health statistics and internal file names, while its wider review found activity involving several Australian government websites.- EndsPublished By: Divya BhatiPublished On: Sep 24, 2026 11:50 IST