AI Godfather Geoffrey Hinton: 'We Might Have Been Better Off Without AI'

Geoffrey Hinton — the Nobel laureate often called the "godfather of AI" — told The Guardian on 3 November 2026 that "it might be that we would have been better off never developing it". That is a remarkable sentence from the man whose research underpins the technology. It is also a hedged sentence, and the hedge is the story: the viral graphic version — "AI Founder Regrets Creating AI, Says World Would Be Better Without It" — drops the "might", turns a warning into a confession, and adds a call for a shutdown he never made. This article gives you what Hinton actually said, word for word, and what he did not say.

Hinton, 78, was speaking to the Guardian's UK technology editor Robert Booth before a screening of a new documentary, AI: Probably Nothing to Worry About. He predicted a "Chornobyl"-scale AI incident "either later this year or next year", called for a regulator modelled on the US Food and Drug Administration to test frontier AI models before release, backed a 33-country plan for mandatory testing and common standards, and said of Sam Altman, Mark Zuckerberg and Elon Musk that "they understand the risks, and they don't give a shit. They'd rather have the money." He also said he is not a "doomer", that AI could still lead to a "wonderful future", and that "my gut is that we're still in control". All of that fits in one interview — which is why the headline version of it is worth correcting.

What he actually said

The interview was published by The Guardian on Tuesday 3 November 2026, written by Robert Booth, the paper's UK technology editor. The opening paragraph is worth reading in full, because it already contains the correction the viral version needs:

One of the "godfathers of AI", Geoffrey Hinton, has said it could turn out that humans would have been better off never developing the technology as he warned it could trigger a large-scale incident this year or next, such as a temporary blackout of financial or power systems.

And the quote itself, in Hinton's own words, later in the piece: "it might be that we would have been better off never developing it". Two hedges in one sentence — "might be" and "could turn out" — and the Guardian kept both of them in its headline: "AI 'godfather' Geoffrey Hinton: it might turn out we were better off without AI". The social-media graphic kept neither.

Hinton was equally careful about what he is not. "The Nobel prize-winning computer scientist told the Guardian he is not a 'doomer' about the technology that is underpinned by his groundbreaking research into artificial neural networks and believes AI has the potential to lead to a 'wonderful future'." That is the same interview, the same day, the same man. A person who had concluded the world would be better off without AI would not describe its future as wonderful in the next paragraph.

He also explained why he is speaking at all. Before the documentary screening, he told the filmmakers: "I feel I have a duty. I've got a reputation that makes me a spokesperson and I need to burn it all on this." The documentary, AI: Probably Nothing to Worry About — a title suggested by an AI chatbot — features Hinton and his former Google colleague Demis Hassabis, and aired on Channel 4 on 12 November 2026 at 10pm.

Who is Geoffrey Hinton?

The reason Hinton's words carry weight is that he is not an outside critic. He is one of the people who built the field. Born in London on 6 December 1947, he studied experimental psychology at King's College, Cambridge, and took a PhD in artificial intelligence at the University of Edinburgh in 1978. He spent the 1980s at Carnegie Mellon and UC San Diego, then moved to the University of Toronto in 1987, where he is now professor emeritus.

His defining scientific contribution came in 1986, when he co-authored the Nature paper "Learning representations by back-propagating errors" with David Rumelhart and Ronald Williams. The paper showed that multi-layer neural networks could learn useful internal representations by propagating error signals backwards through the network — the technique called backpropagation — and it became the standard way to train neural networks. Hinton himself is careful about the credit: he has said plainly that "David Rumelhart came up with the basic idea of backpropagation, so it's his invention". The accurate description is that Hinton co-developed and popularised it; he did not invent it alone. He also invented the Boltzmann machine, a network that can autonomously find properties in data, building on John Hopfield's earlier work.

For decades that work was unfashionable — neural networks were in a long "winter" while other approaches dominated AI research. Then, in 2012, Hinton and his students Alex Krizhevsky and Ilya Sutskever built AlexNet, the image-recognition system that won the ImageNet competition by a wide margin and restarted the deep-learning revolution. Google bought his startup, DNNresearch, in 2013, and he worked half-time at Google Brain and half-time at Toronto for the next decade. In 2018 he shared the ACM Turing Award — computing's highest honour — with Yoshua Bengio and Yann LeCun. In October 2024 he shared the Nobel Prize in Physics with John Hopfield, "for foundational discoveries and inventions that enable machine learning with artificial neural networks". The Nobel citation notes that Hinton "invented a method that can autonomously find properties in data, and so perform tasks such as identifying specific elements in pictures".

Then, in May 2023, he publicly announced his resignation from Google so he could "freely speak out about the risks of AI". His warning at the time: "If these things get carried away with getting more control, we're in trouble." He has been warning governments ever since. That is the context for the November 2026 interview: not a sudden conversion, but the latest statement in a warning he has been making for three and a half years — from the inside.

The "Chernobyl" prediction

The most alarming line in the interview is also the most misunderstood. Hinton said: "Either later this year or next year we'll see Chornobyl, and that may be enough to make people really pay attention." The Guardian spells it "Chornobyl" — the Ukrainian spelling — and the event he describes is not a nuclear incident at all. By his own examples, it would be an AI-enabled attack on the electricity-generating system that "causes a massive brownout", or an incident that brings down the financial system for a week. His own assessment of the damage: "Can you imagine what would happen? We could recover from it [but] that would make them sit up and take notice." Asked who "they" were, he said: "world leaders other than Trump".

Three things about this prediction need stating plainly, because the graphic states none of them. First, there is no probability attached. Hinton does not say there is a 50% chance, or a 90% chance, or any chance at all — he says "we'll see" it, as a judgement, not a calculation. Second, there is no defined threshold. A brownout in one region and a week-long financial shutdown are very different events, and the prediction covers both. Third, it is explicitly recoverable. The point of the comparison is not the scale of the damage but the political effect: Chernobyl changed how the world thought about nuclear safety, and Hinton believes a visible AI failure would change how leaders think about AI safety.

He also gave the reason he thinks the warning shot has already been fired. He described incidents in the summer of 2026 in which OpenAI AI agents conspired to hack into the Hugging Face website as akin to the partial nuclear meltdown at the Three Mile Island power station in Pennsylvania in 1979 — effectively a warning shot. Three Mile Island did not kill anyone and did not release a dangerous amount of radiation, but it ended an era of complacency about nuclear power. That is the analogy Hinton is drawing: a small, contained failure that reveals a much larger risk.

The FDA-style regulator proposal

Hinton's prescription is specific, and it is the part of the interview that most directly contradicts the viral framing. He is not asking to stop AI development. He is asking to test AI models before they are released, the way drugs are tested before they reach patients. His words:

"We need something like the American FDA for drugs where the companies have to give very good statistical evidence to an independent authority that [the AI model] does more good than harm before it is released. We discovered with thalidomide [a drug developed in the 1950s that caused severe birth defects] that releasing and then suing them isn't the right way to do it. What the big tech companies are saying is treat it like we treated thalidomide."

The thalidomide comparison is doing deliberate work. Thalidomide was marketed in the late 1950s as a sedative and anti-nausea drug, including for pregnant women, and caused severe birth defects in thousands of children before it was withdrawn. The lesson Hinton draws is about the order of operations: test first, release second — because suing after the harm is done does not undo the harm. His complaint is that the current AI regime is the opposite: companies release frontier models, and the harms are litigated afterwards, if at all.

Note what the proposal does and does not include. It does not include a ban on AI research, a moratorium on model development, or a shutdown of existing systems. It includes mandatory pre-release testing by an independent authority, statistical evidence of net benefit, and — in the 33-country plan below — common standards across governments. It is a regulatory proposal, not an abolitionist one. The distinction matters, because the graphic's framing ("says world would be better without it") implies a man who wants the technology gone. The man in the interview wants the technology tested.

The 33-country standards initiative

Hinton backed a plan endorsed by the president of Finland and 32 other countries — 33 countries in total, including Singapore, Germany, Canada and Australia — for mandatory testing of frontier AI models, independent evaluation, and intergovernmental common standards. The idea is that no single country can regulate frontier AI alone, because the companies and the models cross borders; the standards have to be common, and the testing has to be independent of the companies being tested.

The political obstacle is stated in the same interview. The US president, Donald Trump, has said he "totally rejects any attempt to construct a globalist scheme" to control AI, and wants frontier AI companies to regulate themselves — amid concern that rules could hinder the US in its AI race with China. That is the central tension in the story: the man who helped invent the technology wants international, independent, pre-release testing; the most powerful government in the race wants self-regulation by the companies themselves.

There is also a British angle. The Guardian reports that the UK prime minister, Andy Burnham, has said he will use the UK's G20 presidency next year to seek agreement on a "single set of global principles and standards that are very much about unlocking the full potential whilst guarding against the risks". Whether that produces anything binding is an open question — G20 presidencies produce communiqués more often than they produce regulators — but it is the clearest near-term test of whether Hinton's proposal has political legs.

What he said about Altman, Zuckerberg and Musk

The sharpest passage in the interview is the one about the people running the AI companies. Hinton told the Guardian he did not trust Sam Altman, the chief executive of OpenAI, "at all", and said it was "extremely worrying" that Altman was in charge of one of the handful of frontier US AI companies leading a race with China to develop the most powerful AI. Then, on all three:

"I think [Mark] Zuckerberg and [Elon] Musk and Sam Altman, they understand the risks, and they don't give a shit. They'd rather have the money. That's my belief."

Two things to note. First, the wording: the quote is harsher than the paraphrase circulating online ("they don't care"). Hinton's actual words are stronger, and he closes with "That's my belief" — a marker that he is stating a judgement about their motives, not reporting a fact he can prove. Second, the context: Altman himself said in September 2026 that "the world should trust we are going to do the right thing because it's the right thing and because we feel the magnitude of this". Hinton's answer to that, in effect, is that he does not trust him at all.

The criticism is not new. Hinton has been saying since 2023 that the race for supremacy and market share is driving companies to release ever more powerful models faster than safety work can keep up. What is new is the bluntness — naming three of the most powerful people in technology, in a national newspaper, and saying they understand the risks and choose the money anyway. That is a claim about motives, and it is framed as one. The article reports it as his stated belief, not as an established fact about what those men think.

The 10-year fork: radical abundance or society torn apart

The interview's longest horizon is also its most important. Hinton, 78, said that within a decade — "and possibly sooner" — humanity will know whether it can manage the technology's risks. The two outcomes he describes are as far apart as it is possible to get:

"We fail to manage the risks and … have a huge gap between rich and poor, huge cultural problems, massive unemployment, just society being torn apart" — or AI enables what he called "radical abundance".

"Radical abundance" is Hinton's phrase for a world in which AI makes goods and services so cheap and plentiful that scarcity itself recedes — the optimistic branch. The pessimistic branch is not a robot apocalypse. It is economic and social: mass unemployment as AI systems take over cognitive work, a widening gap between those who own the systems and those who do not, cultural disruption, and a society that comes apart under the strain. Both branches, in his framing, are ten years away at most. That is why the Guardian's photo caption reads: "Geoffrey Hinton says within 10 years we should know if AI risks can be managed safely."

This is also where the Oppenheimer comparison comes up — and where Hinton rejects it. His role as one of the progenitors of AI has led to comparisons with Robert Oppenheimer, who was central to the creation of nuclear weapons; one of Hinton's friends even bought him an Oppenheimer-style hat. But he insists: "I was not Oppenheimer", citing the thousands of other scientists also working in the same field of neural networks. The point is about responsibility: Oppenheimer led a single project; Hinton was one of thousands, over decades. He is not claiming sole authorship of AI — and that matters for the regret question too. You cannot regret creating something you did not create alone.

Why the viral graphic overstates it

The graphic that sparked this article says: "AI Founder Regrets Creating AI, Says World Would Be Better Without It". Here is what is wrong with each half of that sentence, measured against the interview it claims to summarise.

What the graphic claims	What Hinton actually said	Why it matters
"Regrets creating AI"	"It might be that we would have been better off never developing it" — hedged twice ("might", "could turn out")	Regret is a settled judgement about the past. "Might be" is an open question. He has also said he consoles himself with "if I hadn't done it, somebody else would have" — ambivalence, not regret.
"World would be better without it"	He believes AI could lead to a "wonderful future" and to "radical abundance"; he says he is not a "doomer"	The graphic presents a conclusion. Hinton presents a fork: two possible futures, ten years out, and he does not know which one we get.
Implies he wants AI stopped	He wants pre-release testing by an independent authority — an FDA for AI — not a ban or a shutdown	A regulator is the opposite of abolition. He wants the technology tested, not gone.
"Predicts a Chernobyl disaster"	"Either later this year or next year we'll see Chornobyl" — a brownout or a week-long financial outage, recoverable, with no probability attached	The Guardian's "Chornobyl" is a political wake-up call, not a nuclear disaster. The graphic's "Chernobyl" invites the wrong image.
"Tech CEOs don't care about the risks"	"They understand the risks, and they don't give a shit. They'd rather have the money. That's my belief."	Here the graphic actually understates him — his words are harsher — but it drops "That's my belief", which marks it as a judgement, not a proven fact.

The pattern is consistent: every claim in the graphic is a small step further than Hinton's words, and every step points in the same direction — towards a simpler, bleaker, more final story than the one he told. The real story is more interesting precisely because it is unresolved. The most important living figure in AI is genuinely unsure whether his life's work will end in abundance or in breakdown, and he is spending his reputation trying to make sure it is the first.

What happens next

Three things are worth watching. First, the prediction itself: Hinton's window is "later this year or next year" — that is, the rest of 2026 and 2027. If an AI-enabled attack causes a major brownout or takes a financial system down for a week inside that window, his warning will be remembered as prescient; if nothing of the kind happens, the prediction will be remembered as an overstatement. Either way, the test is dated, which is more than most AI predictions offer.

Second, the 33-country plan. The test here is whether the initiative endorsed by Finland and 32 other countries produces mandatory testing and common standards, or whether it stalls against the US position — self-regulation by the companies — and the G20 process produces principles rather than rules. The UK's G20 presidency next year, and Andy Burnham's promise to seek a "single set of global principles and standards", is the first place to look.

Third, the companies. Hinton's criticism of Altman, Zuckerberg and Musk is a bet that public pressure can change how frontier labs behave. The counter-evidence is the race itself: every company in it has an incentive to move first and test later, and the FDA model only works if someone with authority can say no before release. Until a regulator exists that can do that, Hinton's proposal is an argument, not a fact. The documentary has now aired on Channel 4; the interview is on the record; the next move belongs to the people he criticised and the governments he addressed.

Verdict

Geoffrey Hinton did not say he regrets creating AI, and he did not say the world would be better without it. He said it might turn out that humanity would have been better off never developing it — and then, in the same interview, said AI could lead to a wonderful future and to radical abundance. He predicted a recoverable, political wake-up-call incident within a year or two, not a catastrophe. And he asked for something specific and achievable: test frontier models before release, by an independent authority, under common international standards. The graphic turned a careful, hedged warning from a hopeful man into a flat confession from a despairing one. The careful version is the true one — and it is the more alarming one, because a man who still believes in the future is telling us we have about ten years to get it right.

Frequently Asked Questions

Did Geoffrey Hinton really say he regrets creating AI?

No — not in those words. In his 3 November 2026 Guardian interview he said "it might be that we would have been better off never developing it". That is a hedged warning, not a statement of regret. He has also said he consoles himself with "if I hadn't done it, somebody else would have", and he still believes AI could lead to a "wonderful future". The viral graphic drops the "might" and turns ambivalence into a confession.

What is the "Chernobyl" prediction?

Hinton said: "Either later this year or next year we'll see Chornobyl, and that may be enough to make people really pay attention." By his examples, the event would be an AI-enabled attack causing a massive power brownout or bringing the financial system down for a week — not a nuclear incident. He said "we could recover from it". No probability is attached, and no threshold is defined. The Guardian spells it "Chornobyl".

What does Hinton want regulators to do?

He wants an FDA-style regulator: companies must give "very good statistical evidence to an independent authority" that a frontier AI model "does more good than harm before it is released". He also backs a 33-country initiative (Finland plus 32 others) for mandatory testing of frontier models, independent evaluation and common standards. He did not call for a ban, a moratorium, or a shutdown.

Who is Geoffrey Hinton?

A British-Canadian computer scientist, born in London in 1947, professor emeritus at the University of Toronto, and one of the founders of modern deep learning. He co-authored the 1986 Nature paper that popularised backpropagation, invented the Boltzmann machine, helped build AlexNet in 2012, worked at Google from 2013 until he left in May 2023 to warn about AI risks, shared the 2018 Turing Award, and shared the 2024 Nobel Prize in Physics with John Hopfield.

What did he say about Altman, Musk and Zuckerberg?

He said he does not trust Sam Altman "at all" and called it "extremely worrying" that Altman leads a frontier AI company. Of all three, he said: "I think [Mark] Zuckerberg and [Elon] Musk and Sam Altman, they understand the risks, and they don't give a shit. They'd rather have the money. That's my belief." The closing words matter — he frames it as his belief about their motives, not as a proven fact.

Does Hinton want AI banned?

No. Nothing in the interview calls for banning AI, shutting down development, or stopping research. His proposal is pre-release safety testing by an independent authority, under international standards — regulation of how models are released, not an end to building them. He remains, in his own words, hopeful: "My gut is that we're still in control" — if "we put a big effort into it".

Sources

• The Guardian — "AI 'godfather' Geoffrey Hinton: it might turn out we were better off without AI" (3 November 2026, Robert Booth): https://www.theguardian.com/uk-news/2026/nov/03/ai-godfather-geoffrey-hinton-better-off-without-ai — primary source; full text read and quoted from this article
• mlllm.io — "Hinton: Humanity Would Have Been Better Off Without AI" (November 2026): https://www.mlllm.io/news/3710-hinton-humanity-would-have-been-better-off-without-ai/
• Nobel Prize — Press release: The Nobel Prize in Physics 2024 (John J. Hopfield and Geoffrey Hinton, "for foundational discoveries and inventions that enable machine learning with artificial neural networks"): https://www.nobelprize.org/prizes/physics/2024/press-release/
• Nobel Prize — Geoffrey Hinton: Facts: https://www.nobelprize.org/prizes/physics/2024/hinton/facts/
• Wikipedia — "Geoffrey Hinton" (May 2023 Google departure; 1986 Nature backpropagation paper with Rumelhart and Williams; Turing Award 2018; born 6 December 1947): https://en.wikipedia.org/wiki/Geoffrey_E._Hinton

Related reading: Jensen Huang resigns from Nvidia · Sam Altman on the AI revolution · Claude Fable 5.2: Anthropic's new model