Many serious people say AGI is dangerous

If the idea that advanced AI could wipe out humanity sounds like science fiction or a fringe worry, it is worth knowing who has said it. They include two of the three researchers who shared computing’s highest prize for the methods behind modern AI, and the heads of the companies building it.

None of this proves the risk is real. Experts can be wrong, and some equally serious people disagree (see below). It does show that the worry is mainstream among people who know the technology best. The argument itself is laid out in The case for AI risk, step by step.

Every quote below was checked against the source it links to, in November 2026.

The extinction statement

In May 2023 the Center for AI Safety published a one-sentence Statement on AI Risk:

“Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.”

Its signatories include Geoffrey Hinton and Yoshua Bengio, the CEOs of the three leading labs at the time (Sam Altman of OpenAI, Demis Hassabis of Google DeepMind and Dario Amodei of Anthropic), OpenAI co-founder Ilya Sutskever, DeepMind co-founder Shane Legg, Microsoft CTO Kevin Scott and Bill Gates.

Turing Award winners

The Turing Award is often called the Nobel Prize of computing. Geoffrey Hinton, Yoshua Bengio and Yann LeCun shared the 2018 award for deep learning, the approach behind today’s AI. Hinton and Bengio now warn about extinction risk; LeCun is the most prominent skeptic.

Geoffrey Hinton left Google in May 2023, saying he “left so that I could talk about the dangers of AI without considering how this impacts Google” (X, 2023-05-01). He shared the 2024 Nobel Prize in Physics, and in his Nobel banquet speech (2024-12-10) said:

“There is also a longer term existential threat that will arise when we create digital beings that are more intelligent than ourselves. We have no idea whether we can stay in control.”

Later that month, asked on BBC Radio 4 whether he had changed his estimate of a one-in-ten chance of catastrophe, he answered “Not really, 10% to 20%.” The Guardian reported this as his chance of AI leading to human extinction within three decades (Guardian, 2024-12-27).

Yoshua Bengio defined the danger in How Rogue AIs may Arise (2023-05-22) as an autonomous AI system “that could behave in ways that would be catastrophically harmful to a large fraction of humans, potentially endangering our societies and even our species or the biosphere.” In his FAQ on Catastrophic AI Risks (2023-06-24) he argued that “we should take even a small probability of catastrophic outcomes of superdangerous AI seriously, because of the possibly large magnitude of the impact”.

Other Turing Award winners have signed one of the statements on this page: Joseph Sifakis (2007), Martin Hellman (2015) and Andrew Barto (2024) signed the extinction statement, and Andrew Yao (2000) signed the superintelligence statement.

Heads of the AI labs

The people building the most capable systems have said the same, often before they had a commercial stake in saying the opposite. Why people build AGI despite the danger asks why they build it anyway.

Sam Altman, CEO of OpenAI, in Machine intelligence, part 1 (2015-02-25), months before co-founding OpenAI:

“Development of superhuman machine intelligence (SMI) is probably the greatest threat to the continued existence of humanity.”

In OpenAI’s Governance of superintelligence (2023-05-22), written with Greg Brockman and Ilya Sutskever: “Given the possibility of existential risk, we can’t just be reactive.”

Dario Amodei, CEO of Anthropic, asked for his “p(doom)” at the Axios AI+ DC Summit, put the chance that AI goes “really, really badly” at 25%, and the chance it goes “really, really well” at 75% (Axios, 2025-09-17).

Demis Hassabis, CEO of Google DeepMind and a 2024 Nobel laureate in Chemistry, to the Guardian (2023-10-24): “We must take the risks of AI as seriously as other major global challenges, like climate change.” Of the extinction statement, he said he signed it “because we wanted to give credibility to that being a reasonable thing to discuss.”

Elon Musk, OpenAI co-founder and now head of xAI, on X (2014-08-03): “We need to be super careful with AI. Potentially more dangerous than nukes.” In 2026 he still put “at least some percentage chance of it being an existential threat” (source). His page traces how his position has changed.

Other statements and surveys

• The pause letter (Future of Life Institute, 2023-03-22) called for a six-month pause on training systems more powerful than GPT-4, asking “Should we risk loss of control of our civilization?” Signatories include Bengio, Stuart Russell, Musk and Steve Wozniak.
• The Bletchley Declaration (2023-11-01), signed by the governments at the UK’s AI Safety Summit, including the US and China, warned of “potential for serious, even catastrophic, harm” from the most capable models, including “unintended issues of control relating to alignment with human intent”.
• The superintelligence statement (Future of Life Institute, October 2025) calls for “a prohibition on the development of superintelligence, not lifted before there is (1) broad scientific consensus that it will be done safely and controllably, and (2) strong public buy-in.” Signatories include Hinton, Bengio, Yao, Russell and Wozniak, alongside figures as different as Steve Bannon and Prince Harry.
• A survey of 2,778 AI researchers (Grace et al., 2024) found that “Between 38% and 51% of respondents gave at least a 10% chance to advanced AI leading to outcomes as bad as human extinction.” The median estimate for such outcomes was 5%.

Who disagrees

The skeptics are serious too, and they do not all disagree for the same reason.

Yann LeCun, the third 2018 Turing Award winner then Meta’s chief AI scientist, has called the idea of AI as an existential risk preposterous. In a TIME interview (2024-02-13) he said: “It’s not like one day, we’re going to build some gigantic computer and turn it on, and then the next minute it is going to take over the world. That’s the preposterous scenario.” His view is that capability will arrive gradually and that safety will be engineered along the way. That is the question in Can we learn from AI failures before they are fatal.

Andrew Ng, co-founder of Google Brain, replied to the extinction statement in The Batch (2023-06-07): “I have to admit that I struggle to see how AI could pose any meaningful risk for our extinction.”

Arvind Narayanan and Sayash Kapoor, computer scientists at Princeton, argue in AI as Normal Technology (2025-04-15) that AI will spread slowly through the economy like electricity or the internet: “We view AI as a tool that we can and should remain in control of, and we argue that this goal does not require drastic policy interventions or technical breakthroughs.” This is a dispute about AI takeoff.

Timnit Gebru, Emily Bender and co-authors argue that the focus on future risk distracts from present harms. Replying to the pause letter (DAIR, 2023-03-31), they wrote that “the focus of our concern should not be imaginary ‘powerful digital minds.’”

Two things are worth noticing about the disagreement. First, many skeptics do not say advanced AI would be safe; they say it is far off, will arrive slowly, or can be kept under control, which are claims at particular steps of the argument. Second, people on both sides are AI experts, so the question cannot be settled by deferring to experts. It has to be settled on the arguments.