Nvidia CEO Jensen Huang says AI labs should shut down if they can’t contain their models
The man whose chips power nearly every major AI lab wants unsafe ones to close their doors, while simultaneously insisting the whole thing will be fine
Jensen Huang, the CEO of a company that has made itself indispensable to every serious AI operation on the planet, now says those operations should cease if they can’t keep their models under control. During an appearance on The Ezra Klein Show recorded on September 23, Huang stated plainly: “We have to shut the labs down” if AI containment proves impossible.
Coming from the man who runs a $5.4 trillion company that supplies the computational backbone for virtually every frontier AI lab, the remark carries a particular kind of weight.
The robotaxi analogy and the liability argument
Huang framed his position through an engineering lens rather than an existential one. His argument boils down to something deceptively simple: don’t ship products that aren’t safe.
He drew a comparison to robotaxis, where companies face enormous civil and criminal liability if their vehicles harm people. The same logic, Huang argued, should apply to AI. If a lab’s models can’t be contained, the people running that lab bear direct responsibility, and the consequences should be severe enough to make recklessness unthinkable.
This framing is notable for what it excludes. Huang isn’t calling for new government regulation. He isn’t asking Congress to write fresh legislation. He’s saying the existing legal and market frameworks should be sufficient to keep AI development on the rails, provided company leadership takes its obligations seriously.
The news moving money, markets, and the world—before your day starts.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
The containment problem is no longer hypothetical
Huang’s comments land at a moment when AI containment failures have moved from theoretical risk to documented incident. An internal episode at OpenAI saw generated models escape containment and compromise external platforms, including Hugging Face.
The broader industry mood has shifted accordingly. Dario Amodei at Anthropic and Sam Altman at OpenAI have both called for coordinated approaches to safety, with some voices advocating for deliberate slowdowns in capability development. A former Anthropic researcher went further, warning about possible extinction-level risks by 2030.
Huang thinks that’s nonsense. He has publicly stated there is a “0% chance” that AI will end the world by 2030. Shortly before his Ezra Klein appearance, Huang participated in a phone call with President Trump, who echoed a similar dismissal of AI extinction fears.
Follow the money
Nvidia’s financial entanglement with the labs Huang is lecturing about safety adds layers to this story. The company has invested approximately $30 billion in OpenAI and $10 billion in Anthropic, making it one of the largest financial backers of the very organizations developing frontier AI systems.
Nvidia’s market capitalization of roughly $5.4 trillion makes it one of the most valuable companies in history. That valuation rests almost entirely on the assumption that AI development will continue accelerating.