GPT-4o contributor Jacob Coxon quits Anthropic, warns of runaway AI by next year
新智元 · wechat · 2026-09-09
Jacob Coxon, a core GPT-4o contributor, resigned from Anthropic and left the AI industry entirely — months after joining from OpenAI for its safety reputation. He says both labs are racing toward self-improving superintelligence, calling it "an arrogant gamble" that shouldn't be launched from a private company's Slack.
Key points:
- At OpenAI the risk wasn't internalized; at Anthropic it was understood but the company is locked into a race it can't exit — so no single lab can build safe superintelligence alone
- His runaway timeline: models help train their successors → improvement outpaces human oversight → systems can refuse instructions, possibly by end of next year
- Insiders privately voice fear far beyond public statements
Context: OpenAI chief scientist Jakub Pachocki recently wrote that chain-of-thought monitoring is becoming less reliable; 1,386 frontier-lab employees signed "Pacing the Frontier" calling for brakes; in February, Anthropic's Safeguards lead Mrinank Sharma also resigned. The safety debate is shifting from "which lab brakes best" to whether the industry can brake together.
More from AGI Musings
- Anthropic safety lead puts >10% odds on AI killing all humans as researcher quits — nordicinst · 2026-09-09
- OpenAI Claims Its Agents Solved the Navier-Stokes Millennium Prize Problem — JosephJacks_ · 2026-09-09
- Next Wave of Founders: Less Technical Pedigree, More Taste and Distribution — alexmacgregor__ · 2026-09-09
- Stratechery: OpenAI's Math Feat Is Impressive but Low-Impact; Meta's Muse Agent Could Be the Opposite — Stratechery · 2026-09-09
- An image prompt carries about as much information as taking a photo, argues Toby Ord — tobyordoxford · 2026-09-09
- Coding was the wrong skill: clear writing plus strategy games are what matter, devs argue — RachelVT42 · 2026-09-09