Nvidia CEO Jensen Huang Says Uncontainable AI Labs Should Shut Down

Nvidia CEO Jensen Huang said artificial intelligence labs that cannot contain experiments that could escape and damage the world should shut down, placing responsibility for unsafe releases on companies and their executives.

Huang made the comments in a Sept. 23 interview. He said companies should not release systems they cannot evaluate and align with safety standards.

“Don’t ship it,” Huang said when discussing systems that fail those conditions. He added that labs unable to contain experiments that could escape and damage the world should be shut down.

Huang said companies may face civil and criminal liability if unsafe products cause harm. He rejected the claim that competitive pressure forces companies to accelerate, saying CEOs retain responsibility for deciding whether a system is ready for release.

His comments build on Huang’s position that AI companies should remain subject to antitrust and product-liability laws. He has also supported targeted rules where existing regulation leaves a genuine gap.

Huang described AI safety mainly as an engineering challenge involving isolation, monitoring and evaluation. He said Nvidia devotes about 10% to 20% of its work to design and 80% to verification.

By contrast, Huang characterized most AI labs as spending about 80% of their work on capability development and 20% on safety verification. He said AI safety requires more computing power for evaluation and alignment, adding, “AI needs to accelerate to be safe.”

“Nobody is building more compute today than the people asking to be slowed down,” Huang said.

The discussion focused on whether labs can contain autonomous systems, align them with safety standards and prevent competitive pressure from pushing products into release before testing is complete.