Ajeya Cotra
Senior Program Officer
Safety researchersOne of the most influential grant-makers in AI safety; her public estimate that catastrophe is more likely than not by 2040 anchors the worried end of serious, fund-level analysis.
Voices
Nobody here is a villain in one line. Each entry gives the person's own words and a link to where they said them.
Andrew Ng
Founder of DeepLearning.AI and Google Brain; former Baidu chief scientist; Stanford adjunct professor
SkepticsNg taught millions of people machine learning and founded Google Brain. He thinks extinction fears are baseless—comparable, in his earlier words, to worrying about overpopulation on Mars—and that AGI is decades away. He worries instead that doom rhetoric gives big companies an excuse for regulation that would stifle open-source competitors.
Arvind Narayanan
Computer scientist; co-author of 'AI Snake Oil' and 'AI as Normal Technology' (with Sayash Kapoor)
SkepticsNarayanan argues AI should be treated as a 'normal technology' whose effects unfold over decades through slow institutional change, not as a looming superintelligence. He thinks doom probabilities are unusable for policy and that the best protection is resilience and preventing a few companies from concentrating power.
Aza Raskin
Co-founder
Ethics and justiceInterface designer and ethicist who frames the AI race as a collective-action trap that individual companies cannot exit alone.
Connor Leahy
Co-founder of EleutherAI; former CEO of Conjecture; US Director of ControlAI; co-author of 'The Compendium'
Safety researchersLeahy helped build one of the first open-source GPT clones, then became one of the loudest voices for stopping the race entirely. Through ControlAI he lobbies parliaments for a ban on superintelligence and on AI systems capable of improving themselves.
Dan Hendrycks
Director of the Center for AI Safety; advisor to xAI and Scale AI; co-author of 'Superintelligence Strategy'
Safety researchersHendrycks runs the nonprofit that organised the 2023 one-sentence statement comparing AI risk to pandemics and nuclear war, signed by most lab CEOs. He has said his own probability of doom is above 80%, and with Eric Schmidt he proposed treating superintelligence like nuclear weapons—with deterrence between states and strict control of chips.
Daniel Kokotajlo
Former OpenAI governance researcher; lead author of the 'AI 2027' scenario; executive director of the AI Futures Project
Safety researchersKokotajlo walked away from millions in OpenAI stock rather than sign a gag clause, then wrote 'AI 2027', a widely read scenario of AI automating AI research within a few years. He puts the chance of things going 'horribly wrong' around 70% and now expects superhuman coding around 2028.
Dario Amodei
CEO and co-founder of Anthropic
Lab leadersAmodei left OpenAI to found Anthropic on the premise that frontier AI is dangerous and someone safety-focused should build it. He says there is roughly a one-in-four chance things go 'really, really badly', expects a 'country of geniuses in a datacenter' possibly by 2027, and lobbies for chip export controls and transparency laws rather than a pause.
Demis Hassabis
CEO and co-founder of Google DeepMind; 2024 Nobel Prize in Chemistry
Lab leadersHassabis leads Google's AI lab and won a Nobel for AI that predicts protein structures. He believes human-level AI is roughly five to ten years away and could be ten times bigger than the Industrial Revolution. He does not give a doom number but says the risk is 'non-negligible', wants urgent safety research, and admits international cooperation looks 'very difficult'.
Eliezer Yudkowsky
Founder of the Machine Intelligence Research Institute; co-author of 'If Anyone Builds It, Everyone Dies' (2025)
Safety researchersYudkowsky has argued since the 2000s that a superhuman AI built with today's methods would almost certainly kill everyone. He thinks the six-month pause proposed in 2023 was far too weak and wants a global, enforced halt to frontier AI development. His 2025 book with Nate Soares brought that case to a mass audience.
Elon Musk
CEO of Tesla, SpaceX and xAI; OpenAI co-founder (departed 2018)
Lab leadersMusk has warned for a decade that AI is an existential threat, signed the 2023 pause letter, and then launched his own frontier lab, xAI. He puts the odds of annihilation at around 10-20% and expects AI to be smarter than all humans combined by about 2030—while arguing the upside justifies pressing ahead.
Emily M. Bender
Computational linguist; lead author of 'On the Dangers of Stochastic Parrots'; co-author of 'The AI Con' (2025)
SkepticsBender coined 'stochastic parrots' to describe large language models and argues the whole extinction debate is a distraction manufactured by the same companies profiting from AI. She wants regulators to focus on transparency, accountability and the workers and creators whose data and jobs are being taken now.
Eric Schmidt
Former CEO
Lab leadersThe most prominent accelerationist from big tech: sees AI as a race to be won, and treats safety regulation mainly as a risk of losing it.
Fei-Fei Li
Co-director, Stanford Institute for Human-Centered AI
OtherThe godmother of ImageNet and the leading voice for human-centered AI: argues the technology is made by people and can be steered by people.
Gary Marcus
Cognitive scientist; NYU professor emeritus; author of 'Taming Silicon Valley' (2024); prominent LLM critic
SkepticsMarcus is the best-known critic of the idea that today's chatbots are on the road to human-level intelligence—but he is not a doom-denier. He now puts his own catastrophe odds around 3%, mainly because of reckless deployment by powerful people, and wants an FDA-style agency, liability rules and audits.
Geoffrey Hinton
Deep-learning pioneer; 2018 Turing Award; 2024 Nobel Prize in Physics; left Google in 2023 to speak freely about AI risk
Forefathers of the fieldHinton helped invent the neural-network techniques behind today's AI, then quit Google in 2023 so he could warn about it. He puts the odds that AI wipes out humanity in the next few decades at roughly one in five to one in ten, says superintelligence could arrive within five to twenty years, and argues only government regulation—not market forces—will make companies take safety seriously.
Helen Toner
Director of Strategy at Georgetown's Center for Security and Emerging Technology; former OpenAI board member (2021-2023)
Policy and governmentToner was on the OpenAI board that briefly fired Sam Altman in 2023 and has since become a leading policy voice arguing that self-regulation by AI companies fails 'when money is on the line'. She pushes for transparency rules and whistleblower protections rather than either a pause or laissez-faire.
Holden Karnofsky
Co-founder and former co-CEO
Safety researchersCo-architect of the most-important-century case: that we may live in the hinge of history because AI is likely to transform everything within our lifetimes.
Holly Elmore
Executive Director of PauseAI US; organiser of protests calling for a global pause on frontier AI
OtherElmore leads the U.S. arm of PauseAI, a grassroots movement that pickets AI labs and lobbies legislators for a worldwide halt to training ever-more-powerful systems until they can be shown to be safe. She refuses to put a number on doom, arguing that uncertainty itself is the reason to stop.
Ilya Sutskever
Co-founder and former Chief Scientist of OpenAI; co-founder and CEO of Safe Superintelligence Inc.
Forefathers of the fieldSutskever was the scientific mind behind OpenAI's biggest breakthroughs, took part in the 2023 board revolt over safety, and then left to found a company whose single goal is a 'safe superintelligence'. He has not given a probability of doom, but expects machines that can do everything humans do within roughly five to twenty years and says the transition must be gradual and aimed at AI that cares about sentient life.
Jan Leike
Alignment researcher; former co-lead of OpenAI's Superalignment team (resigned May 2024); now leads alignment at Anthropic
Safety researchersLeike ran OpenAI's team tasked with controlling superhuman AI and quit in 2024 saying safety had 'taken a backseat to shiny products'. He now leads alignment work at Anthropic. He has agreed with a 10-90% range for catastrophe—essentially saying the honest answer is deep uncertainty.
Joy Buolamwini
Founder of the Algorithmic Justice League; author of 'Unmasking AI' (2023); 'Gender Shades' researcher
Ethics and justiceBuolamwini's research showed commercial face-recognition systems failed most on darker-skinned women, and her advocacy helped push several companies out of the police market. She calls AI's victims 'the excoded' and argues that documented harms happening today matter more than hypothetical machine takeovers.
Kate Crawford
Scholar of AI's social and environmental costs; author of 'Atlas of AI' (2021)
Ethics and justiceCrawford's 'Atlas of AI' traces where AI actually comes from: mines, power plants, water, underpaid labour and scraped data. She argues the industry hides these costs and wants laws forcing disclosure of AI's environmental footprint, treating AI as an extractive industry rather than a would-be superintelligence.
Lina Khan
Former Chair
Policy and governmentBrought competition law to the AI era: argues that who owns the infrastructure of intelligence is a democratic question, not just a business one.
Liv Boeree
Science communicator and former poker champion
Safety researchersA leading public educator on existential risk who translates game-theoretic race dynamics into plain language for mass audiences.
Max Tegmark
Physicist; president and co-founder of the Future of Life Institute (organiser of the 2023 pause letter and 2025 superintelligence statement)
Safety researchersTegmark is the MIT physicist whose institute organised the famous 2023 'pause' letter and the 2025 call to ban superintelligence outright. His group's analysis put the chance of losing control of a superintelligence at 90%, and he wants companies forced to publish that number the way Manhattan Project physicists calculated the odds of igniting the atmosphere.
Melanie Mitchell
Computer scientist and complexity researcher; author of 'Artificial Intelligence: A Guide for Thinking Humans'
SkepticsMitchell studies how far today's AI really is from human-like understanding and argues predictions of imminent superintelligence repeat a 70-year pattern of over-promising. She wants regulation aimed at concrete harms—fraud, bias, surveillance—not science-fiction scenarios.
Meredith Whittaker
President of the Signal Foundation; co-founder of the AI Now Institute; former Google researcher and walkout organiser
Ethics and justiceWhittaker runs the encrypted messenger Signal and helped organise the 2018 Google walkout. She argues AI is built on a surveillance business model, that 'doomsday' talk distracts from present harms, and that the newest 'agent' products threaten to hand companies control of everything on your device.
Mustafa Suleyman
CEO of Microsoft AI; co-founder of DeepMind and Inflection; author of 'The Coming Wave'
Lab leadersSuleyman co-founded DeepMind and now runs Microsoft's AI division. He argues the central challenge is 'containment'—keeping ever-smarter systems under control—and says nobody yet has a reassuring answer. He wants AI that never pretends to be a person and a bounded, 'humanist' superintelligence rather than an autonomous one.
Nate Soares
President of MIRI; co-author of 'If Anyone Builds It, Everyone Dies'
Safety researchersSoares runs MIRI, the oldest AI-safety organisation, which in 2023 gave up on solving alignment in time and pivoted to warning the public. He puts the chance of human extinction from AI above 95% and wants an international ban on building superintelligence.
Nick Bostrom
Philosopher; author of 'Superintelligence' (2014) and 'Deep Utopia' (2024); founder of the former Future of Humanity Institute
Safety researchersBostrom's 2014 book 'Superintelligence' put AI extinction risk on the map for Silicon Valley. A decade on he still takes the danger seriously but now warns against over-caution too, arguing that never building superintelligence would itself be a catastrophe and that humanity should 'feel our way through' the transition.
Paul Christiano
Alignment researcher (inventor of RLHF at OpenAI); founder of ARC; Head of AI Safety at the U.S. AI Safety Institute / CAISI (from 2024)
Safety researchersChristiano invented the technique (RLHF) used to make chatbots follow instructions, then left OpenAI to work on safety full-time and now leads safety at the U.S. government's AI standards body. He estimates roughly a one-in-five chance that most humans die within a decade of powerful AI, and close to even odds that humanity irreversibly damages its future.
Robert Miles
AI safety educator
Safety researchersThe most-watched explainer of AI alignment on the internet; his channel introduced millions to why making AI want the right things is unsolved.
Roman Yampolskiy
Computer scientist; director of the Cyber Security Laboratory; author of 'AI: Unexplainable, Unpredictable, Uncontrollable'
Safety researchersYampolskiy is the academic with the most extreme published estimate—essentially certainty that uncontrolled superintelligence would end humanity—because he argues a permanently safe superintelligence is as impossible as a perpetual-motion machine. He wants narrow AI only and a permanent ban on general superintelligence.
Rumman Chowdhury
CEO of Humane Intelligence; former U.S. Science Envoy for AI; former director of Twitter's META team
Policy and governmentChowdhury ran the White House-backed public red-teaming of AI models and served as America's first science envoy for AI. She wants independent testers to be able to probe AI systems the way safety inspectors probe cars, and thinks the real work is fixing harms already happening, not debating machine consciousness.
Sam Altman
CEO and co-founder of OpenAI
Lab leadersAltman runs the company behind ChatGPT and is racing to build superintelligence while saying it is the most dangerous technology humanity has made. He avoids giving a probability of catastrophe, has told Congress the technology 'can go quite wrong', and now says the 'takeoff has started', with fully automated AI researchers targeted for 2028.
Shane Legg
Co-founder and Chief AGI Scientist
Lab leadersThe DeepMind co-founder who coined AGI as a research goal; gives the most-cited concrete timeline from inside a frontier lab: 2028.
Stuart Russell
Computer scientist; co-author of the standard AI textbook; author of 'Human Compatible'; director of the Center for Human-Compatible AI
Forefathers of the fieldRussell literally wrote the textbook on AI and has spent the last decade arguing that the field's standard way of building systems—give them a fixed objective and let them optimise—is unsafe once machines outsmart us. He does not publish a doom percentage, but says unregulated competition is 'Russian roulette with every human being on earth' and wants enforceable international red lines.
Timnit Gebru
Founder of the Distributed AI Research Institute (DAIR); former co-lead of Google's Ethical AI team
Ethics and justiceGebru was pushed out of Google after co-writing a paper on the dangers of large language models, then founded an independent research institute. She argues the extinction narrative is a distraction promoted by the very companies causing today's harms—exploited labour, stolen data, biased systems—and that 'AGI' is too vague to be made safe at all.
Toby Ord
Philosopher; author of 'The Precipice' (2020); senior researcher on AI governance and scaling
Safety researchersOrd is the Oxford philosopher whose 2020 book estimated a one-in-six chance of civilisational catastrophe this century, with unaligned AI the single biggest risk at one in ten. He now argues nobody should be confident about timing—his own median for transformative AI is 2038 with huge error bars—and that governments should hedge against it coming early.
Tristan Harris
Co-founder
Ethics and justiceThe conscience of the attention economy turned AI advocate: argues we are repeating the social-media mistake of deploying first and understanding later.
Yann LeCun
Deep-learning pioneer; 2018 Turing Award; former Meta Chief AI Scientist; founder of AMI Labs (world models)
SkepticsLeCun is one of the three 'godfathers' of deep learning and the most prominent scientist who thinks the extinction debate is misguided. He argues today's chatbots are nowhere near human intelligence, that a sudden AI takeover is 'preposterous', and that open-source development is safer than concentrating power in a few labs. He puts the odds of doom below an asteroid strike—while saying such numbers are meaningless anyway.
Yoshua Bengio
Deep-learning pioneer; 2018 Turing Award; chair of the International AI Safety Report; founder of LawZero
Forefathers of the fieldBengio is one of the scientists who built modern AI and now leads the world's main government-backed scientific assessment of AI risk. He has said there is roughly a one-in-five chance things turn out catastrophic, and in 2025 founded a nonprofit to build AI that understands the world without acting in it. He wants binding international rules and has signed calls to pause superintelligence development.