OpenAI Capabilities Researcher Dan Selsam Publishes Personal Statement on AI Risk
DKokotajlo · x · 2026-09-15
Dan Selsam, a capabilities researcher at OpenAI since 2022, has released a public personal statement on AI risk, shared on his behalf by a former colleague since he has no Twitter account.
- Background: 15+ years in AI — early probabilistic programming work at MIT, early development of the Lean Theorem Prover at Microsoft Research, and one of the first demonstrations of neural networks learning to reason during his Stanford PhD.
- At OpenAI: Nearly five years helping pioneer chain-of-thought optimization on language models, and more recently data-efficient pretraining methods.
- Statement: He says he has become extremely concerned about how far language models have come and the risks of future iterations (the post text is truncated; full statement at the link).
A rare public risk statement from someone on OpenAI's capabilities side.
More from AGI Musings
- Cohere CEO Aidan Gomez: Terminator-style AI takeover talk shouldn't enter public debate — cohere · 2026-09-15
- Stop debating p(doom) numbers — argue over the function that produces them, says Shan Rizvi — ShanRizvi · 2026-09-15
- Thom_Wolf maps why AI leaders who fear AI still race to build it; Manning flags their hubris — agihouse_org · 2026-09-15
- Physicist on AI risk: we build it like 1940s reactors, but with no conservation laws — AryHHAry · 2026-09-15
- Zeynep: real threats come from good-enough AI at scale, not runaway superintelligence — agstrait · 2026-09-15
- From dark matter detectors to voice AI: how the LZ experiment seeded Deepgram — deepgramscott · 2026-09-15