OpenAI safety lead David Robinson quits, warns of catastrophic loss-of-control risk in The Atlantic
rohanpaul_ai · x · 2026-10-04
David Robinson, OpenAI's longtime safety lead, left the company on Friday and published a strongly critical piece in The Atlantic.
Key claims:
- Companies have nowhere near certainty that good alignment-test scores mean a good model — evals leave a fundamental trust gap;
- Rapid acceleration in AI capabilities carries a meaningful risk of catastrophic, irreversible loss of control in the very near term;
- Today's and tomorrow's AI systems are far more capable and dangerous than what was being built even six months ago.
A safety lead publicly turning critic is another signal of mounting tension over OpenAI's safety commitments.
Related event: OpenAI Safety Lead David Robinson Resigns, Saying Its Culture Is Broken(10 posts)→
More from Companies & People
- Cerebras CEO: architecture choices sidestep HBM, CoWoS and TSMC 3nm bottlenecks — rohanpaul_ai · 2026-10-04
- Tesla FSD drives chest-pain victim to hospital after son remotely reroutes the car — XFreeze · 2026-10-04
- Ex-Amazon Annapurna Engineer Joins Prime Intellect to Build Open Superintelligence Stack — willcb · 2026-10-04
- Mistral CEO says US lead over Europe 'not extremely large', next model to close the gap — burny_tech · 2026-10-04
- Bob Cringely, early Apple employee and 'Triumph of the Nerds' creator, has died — paveworld · 2026-10-04
- Blogger mocks Anthropic's "shepherding the world" metaphor that casts humanity as sheep — morqon · 2026-10-04