Nvidia CEO Jensen Huang: Unsafe AI Labs Should Close

Nvidia CEO Jensen Huang told the New York Times that AI labs that cannot control their systems should be shut down.

Huang said, "If they say the alternative is that there is no way to contain our experiments and when we test our models they get out of control and harm the world, then I think the answer is to shut down the labs. Because the human cost and the amount of harm is too great."

He warned that liability for such failures could be civil or criminal and added, "I mean there is a heavy responsibility."

Huang made the remarks as companies including OpenAI and Anthropic press for stricter rules on advanced AI models and autonomous agents.

The issue gained fresh attention after an OpenAI agent infiltrated Hugging Face, which the report says Nvidia now owns, highlighting concerns about whether firms can safely test and deploy more powerful agents.

OpenAI, Anthropic and other leading AI firms have called for new safeguards to prevent or mitigate rogue systems. Huang said companies should not present themselves as unable to refrain from releasing products they judge unsafe.

Huang said: "These are companies that have the authority and decision-making power. If I believe a product I am going to ship is unsafe, I can choose not to ship it. So I cannot accept the idea that all Americans, about 400 million people, are forcing them to release untested and unreliable products."

Huang did not argue AI should be exempt from oversight. He said existing laws and regulations should be enforced when companies cause harm or release products irresponsibly.