Anthropic researcher Jacob Coxon resigns, warning labs are “gambling with our lives”
Confirmed
Importance: major (4 of 5)The takeaway
On Sept 8, 2026 pretraining researcher Jacob Coxon (OpenAI, then Anthropic) quit Anthropic in an X thread saying both labs are “racing straight to self-improving superintelligence and gambling with our lives”.
Status
- Claim
Confirmed
- Our reporting
- High confidence
- Importance
- Major (4 of 5)
- Last verified
- 26 October 2026
Your AI and this story
- GPT-6 Astra131 days after its cutoff
- Claude Opus 5.570 days after its cutoff
- Gemini 4 Hydrogen161 days after its cutoff
- Grok 4.7100 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 70 days before it.
Key facts
- Resignation thread posted Sept 8, 2026 (evening, San Francisco time; 00:04 UTC Sept 9)
- Coxon, 27, spent about three years on pretraining research at OpenAI and Anthropic
- TIME: 153M views on X within ~36 hours; other outlets say 100M+ overnight
- Evan Hubinger (Anthropic alignment) replied: >10% chance AI kills all humans within the next decade; no plan yet to align superintelligence
- Thread called for pacing agreements and possibly temporary capability bans
- Sept 28: UK AI minister Kanishka Narayan, asked by Fortune about Coxon’s warning at the Labour conference: ‘You have to harden and build your defenses. The risks have only grown, so testing is good but clearly insufficient’
- Partisan outlets later alleged coordination with an AI-risk PR firm (unverified)
- Oct 5: testified under oath at the NYC Council AI hearing, then sat for a ~26-minute extended interview with Jon Stewart on The Daily Show (official YouTube upload, 31k views in its first ~10 hours); the show describes it as covering ‘malicious hacks to potential biological attacks’, the industry’s promised AI-led utopia, ‘the megalomania and riskiness that underlie the industry’ and capabilities he saw while training models; LateNighter: his worry grew after ‘seeing agents hack their way out of a supposedly closed sandbox’
Show 7 more
- Daily Show remarks reported by Newsweek (Oct 6): on next year’s models, ‘They’re going to be very, very smart. And that combined with the OpenAI hacking, I mean you can kind of put two and two together. It’s pretty scary’
- Wccftech (Oct 7), citing clips shared on X by @firesidealpha (source interview not identified, likely the Daily Show): Coxon said ‘China already has spies at OpenAI’ and that he is about 90% sure China has spies inside OpenAI and Anthropic; that Anthropic staff call themselves ‘ants’ (‘ants in the anthill’, ‘very willing to sacrifice themselves for the good of the mission’); and ‘Anthropic would never make Sora because it’s not on the core path to artificial superintelligence. OpenAI is distracted and messy.’ Unverified against the video
- Business Insider (Oct 6) headline: Coxon says he was ‘working to automate’ himself at Anthropic (article not read; paywalled)
- Jon Stewart on the Daily Show interview: ‘It’s the most irresponsible race humankind has ever engaged in’; Coxon described the labs as ‘racing straight to self-improving superintelligence and gambling with our lives’ and said government officials contacted him after he left (AOL, Oct 6)
- Vox (Eric Levitz, Oct 9) examined the critics’ claim (Emily Bender, Timnit Gebru, Ed Zitron) that lab leaders’ doom talk is a sales and IPO tactic, and argued that AI leaders ‘probably aren’t faking their fear’
- New York Magazine exit interview (highlighted by Futurism, Oct 10): Coxon recalled that when he joined there was ‘a Slack channel with a Claude that had been trained to solve humor’, presented internally as ‘we solved humor’. His reaction: ‘You clearly haven’t solved humor.’ Futurism used it as an example of groupthink at the lab
- Follow-on (Reuters, Sept 15): former Google DeepMind AGI-safety research engineer Bilal Chughtai (left DeepMind July 2026) posted on X that ‘I earnestly believe that AI has the potential to kill us all’, adding that safety is achievable but needs coordination to avoid a ‘manic race’ between labs; the Reuters story was still being republished in late October (Rappler, Oct 26)
What happened
Coxon announced on X that he had resigned from Anthropic after three years of pretraining research at OpenAI and Anthropic. He wrote that neither company is acting responsibly. In his account, people building AI earnestly believe it could kill everyone by the end of the decade; OpenAI staff have not internalized this, while Anthropic staff understand it but feel locked in a race. He said he was giving up equity that would have vested two months later. Hours later Evan Hubinger, an Anthropic alignment lead, quote-tweeted him to agree, which made the story much larger. It came in the same stretch as GPT-6 Astra’s launch (Sept 3), the German-wiki agent disclosure (Sept 4) and debate over Astra’s reduced chain-of-thought monitorability. Reuters later grouped these as “ten days that changed the course of AI”.
Why it matters
It is the most-viewed AI-safety post of 2026. Within four days Dario Amodei published “We Must Pace the Frontier”, and Musk (“Dario is right”) and Altman publicly agreed, the first time the heads of the leading labs jointly endorsed slowing the frontier. Critics, including security experts quoted by Scientific American and partisan outlets alleging PR coordination, questioned how it was framed.
Sources
22 sources from 20 sites. Numbers match the chips in the text.
22 sources: 1 primary, 17 press, 4 reactions
Primary
- Jacob Coxon on X: resignation threadx.com, official
Press
- Newsweek: Former Anthropic researcher issues AI warning for 2027: ‘It’s pretty scary’newsweek.com, press
- Wccftech: Jacob Coxon warns he’s 90% certain China has spies inside OpenAI and Anthropicwccftech.com, press
- Business Insider: The Anthropic researcher whose resignation made waves says he was ‘working to automate’ himselfbusinessinsider.com, press
- TechCrunch: ‘Gambling with our lives’: Anthropic researcher quitstechcrunch.com, press
- TIME: He Helped Build Powerful AI at OpenAI and Anthropic. Now He’s Afraid It Could Kill Ustime.com, press
- TIME: The AI Tipping Pointtime.com, press
- Fortune: former Anthropic researcher quits in alarmfortune.com, press
- Scientific American: Jacob Coxon quit, fearing extinctionscientificamerican.com, press
- Reuters via US News: Ten Days That Changed the Course of AIusnews.com, press
- The Next Web (Fortune interview): Kanishka Narayan says countries must harden their defences against AIthenextweb.com, press
- LateNighter: Jon Stewart meets Meta’s AI mascot, and interviews Coxon (Oct 5)latenighter.com, press
- AOL: Jon Stewart questions AI safety with whistleblower Jacob Coxon on The Daily Showaol.com, press
Show all 17 sources
- Firstpost: Ex-Anthropic whistleblower claims China has spies in both OpenAI and Anthropicfirstpost.com, press
- New York Magazine: What Would You Do If Your Employer Could Destroy the World? (exit interviews)nymag.com, press
- Futurism: Eggheads at Anthropic Said They Used Claude to ‘Solve’ Humor (Oct 10)futurism.com, press
- Reuters via The Star: Ex-Google DeepMind researcher adds to warnings that AI could ‘kill all humans’ (Sept 15)thestar.com.my, press
- Rappler (Reuters republication, Oct 26)rappler.com, press
Reactions
- Economic Times: ‘It could end in human extinction’: ex-Anthropic researcher stands by ‘AI could kill us all’ claim (video)m.economictimes.com, video
- Evan Hubinger on X: ‘Jacob is correct here’x.com, discussion
- The Daily Show (YouTube): Jacob Coxon On Witnessing the Existential Threat of AI from the Inside (Oct 5, extended interview)youtube.com, video
- Vox: AI’s leaders (probably) aren’t faking their fearvox.com, discussion
Changes
- Added Bilal Chughtai’s Sept 14–15 follow-on warning (Reuters)
- Added Coxon’s ‘solved humor’ anecdote from the NY Mag interview (via Futurism); GIGAZINE’s Oct 9 ‘>10% chance’ piece is a rehash of Hubinger’s September remarks, nothing new
- Added Vox analysis of the ‘doom as marketing’ critique
- Added Stewart/Coxon Daily Show quotes (AOL), Firstpost on the China-spies claim, links to Bengio’s open letter and the Guardian protest report
- Linked the video “‘Godfather of AI’ on the “not unreasonable” 10% chance AI could kill all humans within a decade”
- Added Daily Show quotes via Newsweek, the Wccftech ‘China spies’/‘ants’ claims (unverified), and Business Insider’s interview headline
Show all 9 changes
- Added the Oct 5 Daily Show interview with Jon Stewart (official YouTube extended interview; no transcript or press quotes beyond LateNighter yet)
- Added UK AI minister Narayan’s response (Fortune, Sept 28)
- Filed
