Probes already deployed as first stage of OpenAI monitoring, discussant says
dfrsrchtwts · x · 2026-11-12
In a discussion on misalignment probes, the author notes tricky open questions about how probe systems get trained and stress-tested, but overall judges them a worthwhile tool.
They add a key detail: as they understand it, these probes are currently deployed as the first stage of OpenAI's monitoring system, citing a source.
More from Safety
- Verona Declaration sets principles for generative AI use in computing research — luislamb · 2026-11-12
- Evitable, founded by AI-risk researcher David Krueger, hires a communications associate at $80K-$98K — DavidSKrueger · 2026-11-12
- Studies: Data Centers Cause Just 4% of US Home Rate Hikes as Senator Backs Moratorium — Eliv_nurotic · 2026-11-12
- DC residents urged to call councilmembers before AV legalization window closes for 4 years — AndyMasley · 2026-11-12
- Security team warns: Telegram's new in-message JS runs unsandboxed, don't upgrade — OwariDa · 2026-11-12
- Beff: coordinated takedown is labs preparing regulatory capture — fight for compute next year — beffjezos · 2026-11-12