Third DeepMind Safety Researcher Quits, Warning AI Could Kill Us All
Bilal Chughtai, an AI safety and alignment researcher at Google DeepMind, publicly resigned warning AI has the potential to kill us all, the third such recent departure from the lab.
The safety-exodus story expanded to a second frontier lab on September 15. Per Bloomberg, Bilal Chughtai, an AI Safety and Alignment researcher at Google DeepMind, publicly resigned with a post stating: “I too am extremely concerned by the default trajectory of this technology. I earnestly believe that AI has the potential to kill us all.” He is described as the third AGI-safety researcher at DeepMind to resign recently with similar warnings, and his public phrasing, that humanity may be running out of time to prevent catastrophic outcomes, matches the register of last week’s Anthropic resignation.
The Exodus Is Now Cross-Lab and Coordinated in Message
The pattern is worth stating precisely. On September 9, Anthropic pretraining researcher Jacob Coxon resigned warning labs are “gambling with our lives,” with Anthropic’s alignment science lead publicly endorsing a greater-than-10% decade extinction estimate. Now Chughtai exits Google DeepMind with matching language, as the third similar DeepMind departure. Two labs, multiple researchers, one shared message delivered through public posts rather than internal channels. When the people whose job is alignment conclude the timeline is unsafe, and choose public resignation as the instrument, internal escalation has demonstrably failed.
What the Departures Are Pointing At
The resignations land in the middle of the most consequential safety fortnight on record: Amodei’s pacing proposal with embedded evaluators, the Astra system card conceding degraded monitorability, and OpenAI asking Congress whether coordinated slowdowns are even legal. The resigners are, in effect, living evidence for the pacing argument: the insiders closest to the models believe the trajectory is dangerous, and their exits remove exactly the institutional knowledge that oversight depends on. Watch whether the departures trigger board-level action at either lab, and whether Congress connects the brain-drain timeline to the antitrust-waiver question currently sitting in the FRONTIER Act debate.
Get Insanely Good at AI
The book for developers who want to understand how AI actually works. LLMs, prompt engineering, RAG, AI agents, and production systems.
Keep Reading
How to Deploy Claude Code Auto Mode in Production
Learn how to configure Claude Code's auto mode to run unattended agent workflows, set up defense-in-depth tool guards, and manage the safety classifier.
Altman, Musk, and Hassabis Back Amodei's Plan: All Four Frontier Labs Agree to Pace
Sam Altman pledged OpenAI will adopt independent evaluators with employee-like access, Elon Musk said Dario is right, and Demis Hassabis endorsed the slowdown, marking the first time all four frontier lab chiefs have publicly aligned on pacing.
Paul Christiano Joins OpenAI's Safety Board as the Oversight Debate Peaks
ARC founder and former US AI Safety Institute adviser Paul Christiano has joined the OpenAI Foundation Board and its Safety and Security Committee, weeks after Astra's launch intensified scrutiny of frontier-model oversight.
Anthropic Researcher Quits the AI Industry, and His Lab's Safety Lead Agrees With Him
Pretraining researcher Jacob Coxon resigned from Anthropic citing a race toward self-improving AI, and alignment science lead Evan Hubinger publicly backed him with a greater-than-10% estimate that AI kills all humans within a decade.
OpenAI's Chief Scientist Says No Lab Has Solved Alignment, Calls for Slowdowns
In an essay titled An Alien Mind, OpenAI chief scientist Jakub Pachocki describes AI as grown rather than designed, admits no lab has solved alignment, and calls for voluntary slowdowns and third-party audited safety frameworks.