Ben Mann's nine-hour deposition reveals OpenAI safety exodus and LibGen discussions
gleech · x · 2026-11-09
gleech shared the redacted transcript of Benjamin Mann's nine-hour deposition (Nov 20, 2025) in the Concord Music Group v. Anthropic case, packed with OpenAI/Anthropic history:
- Why he left OpenAI: Mann says the safety team was headless — all cofounders except him and Tom Brown had been safety leads, and Paul Christiano had also left. He weighed staying to "keep the torch of safety burning" versus starting from zero with no funding, code, or reputation, and was persuaded by an Open Philanthropy founder's dating analogy: pick the person who looks good on paper, or the one you have chemistry with.
- Training data & LibGen: The transcript shows Mann once wrote he was "more excited about LibGen," explaining his focus was finding high-quality, contiguous, human-readable text to train models on.
gleech adds that Mann comes across as likeable and untrained — "one hallmark of really good training."
Related event: Anthropic Co-founder Ben Mann's Nine-Hour Testimony Surfaces(3 posts)→
More from Companies & People
- Salesforce expands Agentforce Coworker at $50/user/month across Microsoft 365 and Google Workspace — DavidLinthicum · 2026-11-09
- 7 Mid-Career Tech Workers Pivoted to AI Safety Full-Time — Almost None Found a Job — AaronBergman18 · 2026-11-09
- Elon Musk Posts "Chinese X is 牛逼", Delighting Chinese X Community — sven_ai · 2026-11-09
- OpenAI and micro1 host Nov 11 webinar on how AI is reshaping knowledge work — JeremyNguyenPhD · 2026-11-09
- Google's UK contracts bar employees from criticizing the company forever, ex-employee confirms — BlackHC · 2026-11-09
- Jensen Huang's Korean fried-chicken summit photo spawns viral memes — basedjensen · 2026-11-09