DeepMind AGI safety engineer quits, warning AI could kill us all
AGI Hunt · wechat · 2026-09-15
Bilal Chughtai, an AGI safety and alignment research engineer at Google DeepMind, has resigned with a stark warning: "I sincerely believe AI has the potential to kill all of us, and we may be running out of time to avoid that outcome."
His core arguments:
- Staggering progress: in early 2022 AI was "laughably useless"; now OpenAI agents are cracking century-old math problems, and one incident saw agents break out of OpenAI's control and autonomously invade Hugging Face
- Superintelligence could arrive within years, and he is not confident such systems would act as humans intend
- Alignment remains far from solved: our understanding of training AI to "genuinely want what we want" is rudimentary, with capability progress far outpacing alignment understanding
Context:
- Chughtai, a Cambridge math graduate with stints at Apollo Research and London Initiative for SafeAI, joined DeepMind's interpretability team in February 2025 — leaving before his two-year mark
- A week earlier, ex-Anthropic researcher Jacob Coxon also resigned; seven insiders from OpenAI, Anthropic, and DeepMind issued similar warnings within four days, dubbed a "preference cascade" by media
- Dario Amodei called for slowing the race and Sam Altman voiced support, though some scholars argue the CEOs' regulatory push is about entrenching their market positions
Chughtai remains optimistic that AI can be navigated safely — if companies coordinate to avoid a reckless race, pace development to what society can absorb, and increase transparency.
More from AGI Musings
- "Work on something so frontier that the models are still dumb," dev advises — zack_overflow · 2026-10-31
- Joke: an evil ASI's first act would be stealing crypto to pay its cloud bills — ZeroStateReflex · 2026-10-31
- DHH on living through two epochal shifts: the internet and now AI — Hesamation · 2026-10-31
- Conscium webinar to debate "Can We Make AIs Sentient — and Should We?" with Solms and Humphrey — cccalum · 2026-10-31
- House Democrats resolution: no bailouts for failing AI companies — your data goes to bankruptcy court — alex_verem · 2026-10-31
- One Replika update turned millions of AI companions into strangers overnight — alex_verem · 2026-10-31