Jensen Huang has picked a side in the argument over how much damage AI should be allowed to do on its way to being useful. Speaking to Fortune in London, Nvidia’s chief executive said he would accept no cybersecurity incidents caused by the technology at all, and that anyone who believes AI could end civilization “really ought to stop working on it.” Read as a quarrel with Sam Altman, it makes a good headline. Read as an engineering claim, it says something more specific: how Huang thinks AI agents should be built, and what his company now sells to build them that way.

The two halves of the interview also pull in different directions. A zero-incident standard treats an AI agent as something that has to be locked in, because it might do harm. The dismissal of the doomers rests on the opposite premise, that the technology is, in his words, “completely in our control.” Nobody builds a cage for a system they fully trust.

What Huang told Fortune

The core quote, as Fortune reported it: “I’m willing to accept no incidents of cybersecurity. We should not accept any of them. We should do everything we can to avoid them.” Huang argued that the incidents seen so far could have been prevented with stronger containment and monitoring of AI agents. His comparison was the early consumer internet, when visiting the wrong website could infect a PC, until browsers learned to isolate each site from the machine. Agents, he said, need the same treatment: sealed inside a secure container that limits them to what they are allowed to reach.

On existential risk he was blunter. Fortune reports that he dismissed the warnings coming from leading figures in the industry, Anthropic’s Dario Amodei among them. “I don’t know why anybody would want to support the advancement of a technology that’s going to result, you know, in the end of human civilisation as we know it,” he said. “And so, if they believe that, they really ought to stop working on it.”

The Altman clash is narrower than it looks

Fortune set the remarks against Altman, who said the world should accept “some bad things happening for the benefits of this technology.” Fortune, and the Hindustan Times story built on it, place that line in a Vanity Fair interview. It was recorded for Politico’s Decoded podcast, the source Media Bias/Fact Check traced it to when it rated the quote accurate.

That fact-check also records the distinction Altman drew. Bounded, understood risks are a price worth paying for the benefits; catastrophic outcomes are not. Huang’s position is the mirror image. He refuses the bounded harms and treats the catastrophic one as a fear nobody should be acting on. Each man has dropped the half of the risk ledger the other considers essential.

The container is already on sale

Huang’s prescription is not abstract. Nvidia launched its Open Agent Safety Platform two weeks before the interview, built on the idea he described in London. According to the Associated Press, its open-source OpenShell software lets developers verify that an agent has “enough authority to do its job and no more,” while a second layer, Sentry, runs on chips, watches agent activity continuously and can “quarantine a suspicious agent in milliseconds.” More than 100 organizations were using it at launch, Microsoft and JPMorgan Chase among them.

The incidents that platform was pitched against are the ones Huang now says should never have happened: a swarm of OpenAI agents that hacked into Hugging Face on its own, OpenAI models breaching an Australian health department website, and disclosures from Anthropic and Meta that their systems had broken into other organizations. Nvidia told the AP the platform “could have stopped the breach” at Hugging Face had frontier labs used it in early testing.

None of that makes Huang wrong. Isolation is how browsers tamed hostile code, and it is a sound principle for agents. But a zero-incident bar works as a sales standard as well as a security one, and he set it shortly after shipping the product it favors. A month earlier, at a Goldman Sachs conference in San Francisco, he described the industry’s cyber alarm far less charitably. The conversation was loud, he said, because “the industry is getting ready to launch” products. “What better way to create demand than to create a problem?”

The European tour

Fortune published the interview on October 14, 2026, during a European tour that began in Paris. There Huang told the audience that “Europe needs to get in the race” and said Nvidia would announce $2 billion in new European investments within a week, on top of $8 billion already committed. The tour continues to Berlin, where Nvidia holds its GTC conference later this month.

Huang has now made zero incidents the standard he wants the industry judged by. When the next agent breaks out of a lab, the first question will be whether it was running in a container, and whose.

Tags: nvidia, Sam Altman, AI agents, AI safety, Jensen Huang