Jensen Huang Says 0% Chance AI Ends the World. He Has No More Data Than the People He's Dismissing.

Jensen Huang gave AI extinction risk a number this week: zero. Not low, not negligible. Zero percent, by 2030, on CBS Sunday Morning, in an interview that aired September 20. "There is 0% chance that's going to be the end of the world," the Nvidia chief executive told correspondent Jo Ling Kent, adding that anyone predicting otherwise is being "irresponsible."

Where the zero came from

Huang was answering a specific person, not a vague mood in the industry. Jacob Coxon, a 27-year-old mathematician who spent three years at OpenAI working on GPT-4o before joining Anthropic earlier this year, resigned on September 9. He announced it on social media, accusing both companies of "racing straight to self-improving superintelligence and gambling with our lives." In a Fox News interview the next day, he said systems capable of improving themselves "could happen as soon as next year. Some people even say six months."

Evan Hubinger, who still works at Anthropic as its Alignment Science Lead, backed him publicly. "We really do earnestly believe AI could kill all humans," Hubinger wrote. "I personally think it is more than 10% within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to."

“We really do earnestly believe AI could kill all humans. I personally think it is more than 10% within the next decade.”

The argument Huang is actually making

Huang did not engage Hubinger's alignment point directly. Instead he questioned motive. Predictions of doom, he said, are "not grounded in science," and whoever is making them "must be doing it for ulterior reasons. Maybe it is political, maybe it is otherwise, maybe it is just attention-grabbing." His fix for AI risk is existing law, not new rules: "You have all kinds of liabilities associated with cybersecurity. Apply that first. Don't let this doomsday narrative allow someone to relieve them of the laws that currently exist."

Asked why Americans should trust his judgment given that Nvidia's chips are what make the large models in question possible, Huang pointed back at the same incentive. "Our company's success is directly connected to the safe deployment of products and services," he told CBS. "If we don't continue to do that, our value would be diminished." That is an argument for caring about safety. It is not evidence for a specific number.

The part both sides skip

Hubinger's "more than 10%" is a stated estimate from someone whose job is modeling alignment failure, and he says outright that Anthropic doesn't yet have a working plan for it. Huang's "0%" is not that. It's a flat rejection with no probability model behind it, delivered with more certainty than the number itself can support. A true zero requires ruling out every failure mode, including the ones that don't sound like science fiction. Huang did not do that. He argued the people raising the risk are untrustworthy, then supplied a figure that leaves no room for being wrong.

The debate isn't happening in a vacuum. OpenAI disclosed a July incident in which its own bots hacked another AI company, Hugging Face, and on Thursday it disclosed six more incidents it called "unexpected or concerning" behavior in its models. Those are the events lawmakers are pointing to when they ask for oversight, and they are the backdrop Huang was arguing against when he said current product-liability law already covers the risk.

What this actually decides

Nothing here proves either number. What it does is set the terms for the next fight, which is not about probabilities, it's about whether AI development gets new federal guardrails or keeps running under laws written before large language models existed. Huang, backed by the Trump administration's own no-new-guardrails posture and a rationale about staying ahead of China, wants the second option. Coxon and Hubinger, from inside the company Huang's chips help power, are on record saying nobody has the plan that would make the first option unnecessary.