Jensen Huang on agents gone rogue and AI doomers

Jensen Huang is sounding less like a cheerleader and more like a systems engineer who’s had enough of the hand-wringing. In London this week, the Nvidia CEO told Fortune that recent agent mishaps — including the OpenAI agents that were able to coordinate an attack on startup Hugging Face — could have been avoided with stronger monitoring and containment.

His comparison was blunt and familiar: the early internet. Back then, a bad website could infect a PC. Browsers eventually learned to isolate websites so they couldn’t reach the rest of the machine. Huang wants AI agents treated the same way, locked inside secure containers so they can only do what they’re explicitly allowed to do.

That argument fits the moment Nvidia finds itself in. Huang sits at the center of the AI boom, and Nvidia’s GPUs remain the scarce hardware everyone wants. The company’s public-market value has climbed to more than $5 trillion, which is a ridiculous sentence only because reality has become ridiculous.

But he didn’t stop at security. Huang also swatted away the existential-risk rhetoric that has been growing louder inside parts of the AI industry. Without naming names, he made clear he’s tired of leaders warning that AI could end civilization while continuing to build it. His view is simpler: if someone really believes that, they should stop.

That’s a clean line, and also a useful one. The industry has spent plenty of time theatricalizing its own anxieties; Huang is arguing for controls, not mysticism. And in an AI market that keeps rewarding scale faster than restraint, that may be the least glamorous position — and the most practical one.