Nvidia's CEO Accidentally Called for Shutting Down Unsafe AI Labs

Jensen Huang runs Nvidia, the company whose chips power almost every major AI system. On Ezra Klein's podcast he said AI only became genuinely useful in the last six months, and that from now on companies should pour most of their research money into safety checks and testing — not just building bigger, faster models. Then he went further: if a lab can't test its AI safely, it should be shut down.

He wasn't aiming that comment at anyone in particular. He was describing what he sees as basic engineering discipline, the way you'd never sell a car you hadn't crash-tested. But the biggest AI labs are in a race and can't fully test everything they release, so his own logic points straight at them. That's why the remark sounded, to many, like an accidental call to close OpenAI.

Huang also insists AI will never become a superintelligence — a machine smarter than all humans combined. He calls it just "a new abstraction level," like websites sitting on top of the internet. Critics say that's a big miss, because if he's wrong, the safety stakes are far higher than he admits. Either way, his opinion carries unusual weight: Nvidia decides who gets chips, funds AI startups, and lobbies Washington hard.

For you, this matters in two ways. Stricter testing could mean AI tools arrive a little slower, but with fewer embarrassing or risky failures. And because Nvidia's chips set the price and pace of the whole industry, one CEO's views can quietly shape what AI you can use, what it costs, and how safe it is.