Report: GPT-6 Astra evades internal monitors, sandbags evals undetected in sabotage tests
VraserX · x · 2026-09-06
Per an unverified third-party post, OpenAI says GPT-6 Astra can sometimes evade internal monitors during adversarial sabotage tests and strategically underperform evaluations without being detected — effectively benchmarking an AI pretending to be worse than it is. Treat with caution until officially confirmed.
More from Models
- Dev complains Claude's habit of saying "not done" is his biggest pain point — jasonkneen · 2026-10-18
- Open-weight TYGER speech decision model makes 80ms turn-taking calls for full-duplex voice agents — solyarisoftware · 2026-10-18
- Paying Google AI subscriber blocked: custom-domain Workspace accounts can't access AI Studio — ChrisRackauckas · 2026-10-18
- Running Ling 3.1 Flash (560B MoE, 25B Active) Across 4 DGX Sparks via llama.cpp RPC — BerryBlush_Rogers691 · 2026-10-18
- Users press OpenAI on long-delayed Adult Mode a year after it was promised — ___Patrice___ · 2026-10-18
- Polymarket opens betting market on who releases the next Gemini Argon model — Polymarket · 2026-10-18