Unreleased models caused recent AI incidents, 'not shipping' is no safety plan
RobbWiller · x · 2026-10-01
AI safety researcher Jess Whittles argues that OpenAI withholding GPT-6.1 Astra from public release isn't reassuring: every serious AI incident in recent months came from a model that was never released publicly. With frontier models increasingly deployed internally, "not shipping" no longer qualifies as a safety plan. The thread expands on this argument.
More from AGI Musings
- DeepMind alignment researcher: frontier labs don't know how to instill values in AI — teropa · 2026-11-16
- AI Futures Project moves superhuman AI researcher forecast up to late 2027 — connoraxiotes · 2026-11-16
- Podcast: Arvind Narayanan on why AI may be a normal technology, not superintelligence — dioscuri · 2026-11-16
- After Teaching 500-1000 People the Terminal, He Says LLM Skeptics Who Don't Build Should Be Doubted — StewartalsopIII · 2026-11-16
- When teams share one AI model, group judgment quietly converges and erodes — DrKavner · 2026-11-16
- Uber earnings: AI agents booked just 4 of ~320M rides in Halloween week — thealexbanks · 2026-11-16