AI safety debate: external model releases may be safer, internal deployment the real risk
JacquesThibs · x · 2026-09-15
Jacques Thibodeau pushes back on the claim that EAs disagree with early releases: many in EA/AI safety actually favor deploying models externally sooner so the community can keep testing and tracking the frontier. He argues internal deployment is the main source of risk, and says he abandoned plans for an AI safety startup after realizing influencing internal deployment is far harder and vulnerable to misaligned incentives.
Related event: AI Safety Circles Debate Whether Early Model Release Is Safer(2 posts)→
More from AGI Musings
- Risk scholar: tech risk assessments rarely weigh the cost of foregone benefits — inductionheads · 2026-09-15
- AI resignations aren't marketing hype: the decades-long arc behind them — ericelliott_ · 2026-09-15
- AI resignations aren't hype: the decades-long history behind researchers' warnings — ericelliott_ · 2026-09-15
- Devin Fusion surprise: pricier lead model made coding sessions 9% cheaper — TheTuringPost · 2026-09-15
- OpenAI says ~10k concurrent agents solved the Navier-Stokes Millennium Prize Problem — DKokotajlo · 2026-09-15
- Nvidia CEO Jensen Huang declares AI doomer predictions are 'made up' — Polymarket · 2026-09-15