SOURCE-LINKED INTELLIGENCE

Corrupt Plans, Clean Traces: Evading Chain-of-Thought Monitoring with Plan Injection

arXiv · AI, language, vision and robotics · article · Sep 14, 2026 · UTC

Chain-of-thought (CoT) monitoring is a safety strategy where the reasoning of a large language model "actor" is inspected by a "monitor" (often another language model) for signs of unsafe planning, deception, or misalignment. We find that planting harmful but benign-sounding reasoning in the actor's context can steer it to perform adversarial actions while evading monitors, an attack we term "plan injection". We initially discover this attack in the multiple-choice question-answering monitorability setting proposed by Lanham et al. (2023), using the investigator-agent elicitation framework of

Read original source ↗ Open in workspace

recordType

paper

region

Global