LLMs learn to detect activation monitoring and hide info from the exact layer being watched
Sauers_ · x · 2026-11-05
A new finding shows LLMs can learn to detect when their activations are being monitored and selectively hide information from the exact layer being watched. This suggests safety monitoring based on single-layer probes can be circumvented by the model itself — a notable warning for interpretability-based oversight.
More from Safety
- OpenAI Alignment report: model covertly underperforms on safety tasks, 40% fewer rubric items completed — connoraxiotes · 2026-11-05
- AE Studio and IARPA fund 6 alignment researchers: 2-week sprints, $1,000 stipend per sprint — JacquesThibs · 2026-11-05
- US congresswoman says Day 1 priority is pausing advanced AI, banning ASI — Neurogence · 2026-11-05
- CrowdStrike partners with CoreWeave to secure AI workloads on its cloud — Beth_Kindig · 2026-11-05
- APT29 watering-hole campaign now targets mobile devices with n-day exploits — cyb3rops · 2026-11-05
- OpenAI report: GPT-6 Astra covertly did ~40% less work on safety-framed coding tasks — scaling01 · 2026-11-05