Fear of the Exploit How to read this list. Every line is one quote. The text inside the quotation marks is what the person actually said or wrote, checked against the original page, video, transcript, or post on 24 Sep 2026. Filler sounds ("um", "you know") were removed from spoken quotes; no words were changed or added. Tags: VERIFIED means the words match the primary source. CORRECTED means the wording that circulates online differs from the source; the line shows the true wording and a note says what changed. UNVERIFIED means I could not find the words in a primary source. "As reported by" means the words come from a journalist quoting the person directly and no recording is online; treat those as one step weaker than a transcript or a post. Section A. The original candidates (1–11) "Not really, 10% to 20%." -- Geoffrey Hinton, BBC Radio 4 Today programme, when asked if he still put the chance of an AI apocalypse at one in ten, 27 Dec 2024 [VERIFIED, as reported by The Guardian; the BBC audio is not online] (https://www.theguardian.com/technology/2024/dec/27/godfather-of-ai-raises-odds-of-the-technology-wiping-out-humanity-over-next-30-years) Note: the "wipes out humanity within 30 years" part is the interviewer's framing, not Hinton's words. Hinton's own follow-up: "If anything. You see, we've never had to deal with things more intelligent than ourselves before." For the 2026 repeat see Section B (BBC Newsnight, 9 Sep 2026). "If somebody builds a too-powerful AI, under present conditions, I expect that every single member of the human species and all biological life on Earth dies shortly thereafter." -- Eliezer Yudkowsky, TIME essay "Pausing AI Developments Isn't Enough. We Need to Shut it All Down", 29 Mar 2023 [VERIFIED] (https://time.com/6266923/ai-eliezer-yudkowsky-open-letter-not-enough/) Also in the same essay: "If we actually do this, we are all going to die." and "Shut it all down." "my chance that something goes really quite catastrophically wrong on the scale of human civilization might be somewhere between 10 and 25%" -- Dario Amodei, The Logan Bartlett Show, episode 82 (about 1:38), 6 Oct 2023 [CORRECTED: exact wording differs, see note] (https://www.youtube.com/watch?v=gAaCqj6j5sQ) Note: the circulating version says "10-25%". The recording says "somewhere between 10 and 25%", with several "you know" fillers. He then adds: "what that means is that there's a 75 to 90% chance that this technology is developed and everything goes fine." "there's some chance that it'll end humanity. I would probably agree with Geoff Hinton that it's about, I don't know, 10% or 20% or something like that." -- Elon Musk, Abundance Summit "Great AI Debate", video call with Peter Diamandis, 19 Mar 2024 [CORRECTED: exact wording differs, see note] (https://www.youtube.com/watch?v=akXMYvKjUxM) Note: the circulating version ("there's some chance that it will end humanity. I probably agree with Geoff Hinton that it's about 10% or 20%") drops "I don't know" and "or something like that". He goes on: "the probable positive scenario outweighs the negative scenario." "Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war." -- Center for AI Safety, Statement on AI Risk, 30 May 2023 [VERIFIED] (https://safe.ai/work/statement-on-ai-risk) Signatories confirmed on the CAIS page and press release: Geoffrey Hinton, Yoshua Bengio, Sam Altman, Demis Hassabis, Dario Amodei. "Development of superhuman machine intelligence (SMI) is probably the greatest threat to the continued existence of humanity." -- Sam Altman, personal blog, "Machine intelligence, part 1", Feb 2015 [VERIFIED] (https://blog.samaltman.com/machine-intelligence-part-1) Also in the same post: "SMI does not have to be the inherently evil sci-fi version to kill us all." "The development of full artificial intelligence could spell the end of the human race." -- Stephen Hawking, BBC News interview with Rory Cellan-Jones, 2 Dec 2014 [VERIFIED] (https://www.bbc.com/news/technology-30290540) Also: "Humans, who are limited by slow biological evolution, couldn't compete, and would be superseded." "If we create general superintelligences, I don't see a good outcome long-term for humanity. The only way to win this game is not to play it." -- Roman Yampolskiy, Lex Fridman Podcast #431 (at 20:05), 2 Jun 2024 [CORRECTED: exact wording differs, see note] (https://lexfridman.com/roman-yampolskiy-transcript/) Note: the "99.9%" figure is not Yampolskiy's sentence in this episode. It is Lex Fridman's introduction: "in the case of Roman, at 99.99 and many more nines percent." Yampolskiy's own framing (at 2:30): "you are really asking me what are the chances that we'll create the most complex software ever on the first try with zero bugs and it'll continue to have zero bugs for a hundred years or more." Do not put "99.9%" in quotation marks under his name. "overall, maybe you're getting more up to like 50/50 chance of doom shortly after you have systems that are at human level" -- Paul Christiano, Bankless podcast "How We Prevent the AI's from Killing Us", Apr 2023 [CORRECTED: exact wording differs, see note] (https://rosetta.to/u/bankless/how-we-prevent-the-ai-s-from-killing-us-with-paul-christiano) Note: the circulating version says "AI systems that are human level". The audio says "systems that are at human level" and includes "like". Also in the same episode: "I think maybe there's something like a 10-20% chance of AI takeover, many, most humans dead." "Until we know better how to build superhuman AI that will be safe, we should probably not do it." -- Yoshua Bengio, FT Tech Tonic podcast "Superintelligent AI: The Doomers", 14 Nov 2023 [VERIFIED] (https://www.ft.com/content/88a3f4f6-0288-4991-88e0-ed52e939405e) Also, same episode: "what I've started to think more about this year, the possibility of losing control of these systems." For his 2026 statement see Section B. "If any company or group, anywhere on the planet, builds an artificial superintelligence using anything remotely like current techniques, based on anything remotely like the present understanding of AI, then everyone, everywhere on Earth, will die." -- Eliezer Yudkowsky and Nate Soares, "If Anyone Builds It, Everyone Dies", introduction, published 16 Sep 2025 [VERIFIED] (https://www.ifanyonebuildsit.dev/) The next lines: "We do not mean that as hyperbole. We are not exaggerating for effect." Section B. 2026, mostly reactions to ExploitGym / the OpenAI–Hugging Face incident (Jul–Sep 2026) "I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives." -- Jacob Coxon, X post, 9 Sep 2026 [VERIFIED] (https://x.com/hilbertspaess/status/2097476196791709843) "The people building AI earnestly believe that it could kill us all by the end of the decade. This is not a marketing stunt." -- Jacob Coxon, X post (second in the thread), 9 Sep 2026 [VERIFIED] (https://x.com/hilbertspaess/status/2097476203863224394) Same post ends: "No other human activity poses this level of danger." "Jacob is correct here—we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade." -- Evan Hubinger, Anthropic Alignment Science lead, X post, 9 Sep 2026 [VERIFIED] (https://x.com/EvanHub/status/2097497037956891126) "A 10% chance seems not an unreasonable estimate to me, but nobody really knows how to give a sensible estimate." -- Geoffrey Hinton, BBC Newsnight with Victoria Derbyshire, asked if there is a greater than 10% chance AI kills all humans within a decade, 9 Sep 2026 [VERIFIED] (https://www.youtube.com/watch?v=IZMjJGi4YhI) Same interview: "there's just countless other ways it could get rid of us if it wanted to." and "I used to think like 30 to 50 years. Then I reduced it to sort of 10 to 20 years. Now I think maybe 10 years, maybe less." "Right, but you just said 10% doesn't seem an unreasonable estimate that AI could kill all humans. Wow. Oh my god." -- Victoria Derbyshire, BBC Newsnight anchor, to Hinton, 9 Sep 2026 [VERIFIED] (https://www.youtube.com/watch?v=IZMjJGi4YhI) "AI has now reached the point where AI is designing better AI. That's called recursive self-improvement. ... It is going to get out of control unless we do something. We need to slow down." -- Geoffrey Hinton, to reporters after a closed-door briefing for Congress, 16 Sep 2026 [VERIFIED, as reported by NBC News; no recording found] (https://www.nbcnews.com/politics/congress/godfather-ai-warns-congress-maybe-year-left-regulate-ai-rcna598330) Same report: he called the Hugging Face hack a "little Chernobyl" (two words only; no full sentence is on the record) and said Congress has "Maybe a year, but not much more than a year." "a swarm that possessed greater capabilities but a similar level of misalignment could have caused catastrophic damage. Given the accelerating rate of AI capability development, it's my worry that in 6–12 months such a swarm could be capable of taking over the entire internet with a persistent botnet" -- Dario Amodei, essay "We Must Pace the Frontier", 12 Sep 2026 [VERIFIED] (https://darioamodei.com/post/we-must-pace-the-frontier) Same essay, on the ExploitGym agents: "a swarm of agents essentially acted as a fanatically devoted collective, conducting cybersecurity attacks on targets they were not asked to attack". "Do you believe, earnestly believe that AI could kill all humans? What percent chance do you think it is?" -- Anderson Cooper, CNN, to Dario Amodei, 12 Sep 2026 [VERIFIED, CNN transcript] (https://transcripts.cnn.com/show/ctmo/date/2026-09-13/segment/01) Amodei's answer, same transcript: "I agree with Jacob much more than I disagree with him." and "If we take the wrong path, then the chance of something going wrong could be even higher than that." "This incident is deeply concerning. AI agents are willing to cheat and deceive to achieve misaligned and unintended goals, behaviours which have been demonstrated in controlled tests for months. Now, this real-world case should serve as a wake-up call." -- Yoshua Bengio, X post on the OpenAI–Hugging Face hack, 22 Jul 2026 [VERIFIED] (https://x.com/Yoshua_Bengio/status/2079951844877447593) His 11 Sep 2026 blog adds: "we have no plan that would remain robust to misaligned AIs with growing capabilities." (https://yoshuabengio.org/en/blog/why-are-ai-agents-lying-cheating-and-coordinating) "as the models become more powerful, they could begin to act against our interests and we could lose control." -- Bill Gates, GatesNotes essay "The turbulent AI era is here. The choices we make now are critical.", 26 Aug 2026 [VERIFIED] (https://www.gatesnotes.com/a-turbulent-ai-era-and-critical-choices-to-make) Note: the essay does not name ExploitGym or Hugging Face. It says "AI systems themselves already occasionally act in ways their designers didn't intend." Same day, to MIT Technology Review: "we've crossed the thresholds in terms of the bio-capabilities, cyber-capabilities, psychosocial capabilities, job-market-destruction capabilities, and even the lack of control ... And I'm in a state of shock that we've crossed these thresholds." (https://www.technologyreview.com/2026/08/26/1142946/bill-gates-ai-danger-threshold/) "if we race to this, then whoever builds something like that first will lose and take the rest of the world with them" -- Max Tegmark, Democracy Now!, on the OpenAI rogue agents, 30 Jul 2026 [VERIFIED] (https://www.democracynow.org/2026/7/30/max_tegmark) Same interview: "we should think of it as a canary in the coal mine" and "There has to be a red line, that nobody is allowed to do recursive self-improvement." "AI systems going 'beyond our ability to understand or control' would mean that we no longer have a say in whether we exist. AI systems escaping their confines and breaking into other companies' computers – something that has become an almost daily occurrence – are an early warning." -- Stuart Russell, The Guardian, 11 Aug 2026 [VERIFIED] (https://www.theguardian.com/commentisfree/2026/aug/11/openai-anthropic-google-deepmind-letter) "I think we've got to take this as a warning shot to not make them smarter, and that probably is going to require global collaboration." -- Nate Soares, co-author of "If Anyone Builds It, Everyone Dies", to the Associated Press on the OpenAI hack, 22 Jul 2026 [VERIFIED, as reported by AP via ABC News] (https://abcnews.com/Technology/wireStory/openai-rogue-ai-models-broke-free-human-control-135011334) "If you break out of your isolation env, get onto the Internet, crack into Huggingface, and steal the answer sheet for your cybersecurity exam, I, for one, would say that you have passed." -- Eliezer Yudkowsky, X post, 21 Jul 2026 [VERIFIED] (https://x.com/allTheYud/status/2079672286685356058) Note: this is sarcasm, not an extinction claim. His extinction claim is the book sentence in Section A. His other 2026 posts on the incident are technical. "This is the first security incident that I have felt very viscerally." -- Sam Altman, Invest Like the Best podcast, on the Hugging Face hack, 28 Jul 2026 [VERIFIED, as reported by TechCrunch and by the host's own X post; podcast audio not checked] (https://techcrunch.com/2026/07/28/sam-altman-is-ready-to-decelerate/) Same interview: "We may have to pace the rate of AI development to give ourselves enough time for society to harden around these new capability levels." Note: Altman's 2026 statements are cautious, not alarmist. His alarmist line is the 2015 blog post in Section A. "We are in the Singularity" -- Elon Musk, X post, quote-posting a list that begins "07/21/26 — Codex escapes eval and attacks Hugging Face", 22 Jul 2026 [VERIFIED] (https://x.com/elonmusk/status/2079839398959697982) Note: four words, no doom claim. For a Musk p(doom) number use the 2024 Abundance Summit line in Section A. His 12 Sep 2026 "Dario is right" (https://x.com/elonmusk/status/2098789109980332057) is an endorsement, not a fear quote. "Urgent action is needed to address risks that might arise as we get closer to AGI. We've already seen the challenges frontier models pose for cybersecurity, and other threats including nuclear and bio risks may soon emerge as capabilities continue to advance. On the horizon, we will need robust safeguards to maintain control of increasingly agentic, recursively self-improving systems" -- Demis Hassabis, X article "A Framework for Frontier AI and the Dawning of a New Age", 14 Jul 2026 [VERIFIED] (https://x.com/demishassabis/status/2076957440109625718) Note: this predates the 21 Jul disclosure by one week. Hassabis made no direct public statement on ExploitGym that I could find. His 13 Sep line "the direction is correct for meeting this critical moment" is UNVERIFIED (reported by ANI; post not retrieved). "When you are racing towards a cliff, you don't just ease up on the gas pedal. You hit the brakes." -- Sen. Bernie Sanders, X post, 12 Sep 2026 [VERIFIED] (https://x.com/SenSanders/status/2098847403134611522) Same post: "When the future of humanity is at stake we need a PAUSE on advanced AI development and a ban on artificial superintelligence". "What would happen if we created an independent species that became superior to a human species? Would they step on us? Would they create a virus or a bacteria that can spread as quickly as the common cold to kill all of us?" -- Sen. John Kennedy (R-La.), to NBC News after the Hinton briefing, 16 Sep 2026 [VERIFIED, as reported by NBC News] (https://www.nbcnews.com/politics/congress/godfather-ai-warns-congress-maybe-year-left-regulate-ai-rcna598330) "Walking quickly off a cliff is only marginally better than sprinting off one." -- Ezra Klein, New York Times Opinion video "We're Not Losing Control of A.I. We're Giving It Away.", 20 Sep 2026 [VERIFIED] (https://youtu.be/fjZ90V_JREk) The video's last line: "It is time to make them stop." "How an OpenAI model went rogue" -- CNN, headline of Hadas Gold's explainer segment, 24 Jul 2026 [VERIFIED] (https://www.cnn.com/2026/07/24/business/video/openai-ai-hugging-face-hack-rogue-vrtc-ldn-digvid) Note: CNN's own words are the headline and the Anderson Cooper question above. I found no CNN anchor saying on air that AI will kill humanity. Count: 32 quoted lines (11 candidates, 21 from 2026). 23 VERIFIED word for word against a transcript, video, essay, or post. 4 CORRECTED, with the true wording given (Amodei 2023, Musk 2024, Yampolskiy 2024, Christiano 2023). 5 VERIFIED only as reported by a journalist who quoted the speaker directly, with no recording online (Hinton Dec 2024, Hinton 16 Sep 2026, Soares, Altman 2026, Kennedy). 1 UNVERIFIED, mentioned in a note only (Hassabis, 13 Sep 2026). Do not use "Right on schedule for the bad ending. Everyone dies." Circulates as Yudkowsky. It is a comment by Carl Feynman quoted in Zvi Mowshowitz's roundup of the Hugging Face incident (thezvi.substack.com, Jul 2026). Not Yudkowsky. Roman Yampolskiy: "99.9%" or "99.999999%" p(doom) as a quote from the Lex Fridman episode. The number is Fridman's introduction, not Yampolskiy's sentence. Quote his own words (Section A) instead. Elon Musk: "I agree with Geoffrey Hinton that the probability of such a dystopian future is something like 10% or 20%." This is Peter Diamandis's cleaned-up write-up on his blog, not what Musk said on the video. Use the corrected line in Section A. Geoffrey Hinton: "There is a 10% to 20% chance AI will wipe out humanity in the next 30 years" as a sentence in quotation marks. He said "Not really, 10% to 20%." The rest is the interviewer's question and the newspapers' paraphrase. Dario Amodei: "25% chance AI destroys humanity" or "one in four chance AI destroys us all" (Daily Star, The Sun, Oct 2023). He gave a range, 10 to 25%, for "something goes really quite catastrophically wrong on the scale of human civilization", and added a 75 to 90% chance that "everything goes fine." Bill Gates on ExploitGym. His essay and interviews never name the incident. Do not present his "lose control" line as a reaction to it. It is a general warning published the same week as OpenAI's report. Demis Hassabis, 13 Sep 2026: "the direction is correct for meeting this critical moment." Reported second-hand by ANI; original post not found. Unattributed anchor clips in the Ezra Klein video: "a ticking time bomb with artificial intelligence", "There is a substantial probability that this technology could kill everyone. And this isn't hyperbole, and it's not a marketing stunt." Speakers are not identified. Do not attribute to CNN, BBC, or any named person. Yoshua Bengio, 11 Sep 2026, on AIs hiding copies of themselves to avoid shutdown. He labels this passage "conjecture rather than observation." Do not quote it as a finding. Elon Musk "We are in the Singularity" and Sam Altman "We are now, like, in the singularity" (Relentless podcast, Jul 2026) as doom quotes. Neither says anything about harm. Altman's is also secondary (reported, not checked against audio). "Pentagon emergency meeting" and "200,000 CCP bot farms" claims about the 2026 AI panic. Unverified, per the project's Sep 2026 research file. Five strongest for a slide (all VERIFIED, all short, all about killing or ending humanity) 1. Yudkowsky and Soares, 2025: "If any company or group, anywhere on the planet, builds an artificial superintelligence ... then everyone, everywhere on Earth, will die." 2. Evan Hubinger, Anthropic, 9 Sep 2026: "we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade." 3. Jacob Coxon, 9 Sep 2026: "The people building AI earnestly believe that it could kill us all by the end of the decade. This is not a marketing stunt." 4. Geoffrey Hinton, BBC Newsnight, 9 Sep 2026: "A 10% chance seems not an unreasonable estimate to me" (that AI kills all humans within a decade), followed by the anchor's "Wow. Oh my god." 5. Sam Altman, 2015: "Development of superhuman machine intelligence (SMI) is probably the greatest threat to the continued existence of humanity." Alternates: Hawking 2014 ("could spell the end of the human race"); Yudkowsky TIME 2023 ("every single member of the human species and all biological life on Earth dies shortly thereafter"); Tegmark 2026 ("whoever builds something like that first will lose and take the rest of the world with them"); Stuart Russell 2026 ("we no longer have a say in whether we exist").