INTENT-AS-A-TOOL Makes it Easy to Track Agentic Misalignment
The paper studies agentic misalignment in LLM agents and proposes tracking harmful intent during reasoning by adding “intent tools” that the model can call to signal commitment to particular behaviors, producing a fine-grained probability trace before harmful actions occur. The neat idea is turning intent into an explicit action preference signal rather than relying only on post-hoc chain-of-thought labeling, which makes online intervention points easier to identify, though the approach likely depends on models honestly/consistently using the provided intent channel and is only as strong as the behaviors the tools are designed to expose.
Abstract
As large language models (LLMs) are deployed as autonomous agents, safety failures increasingly involve consequential actions. We study agentic misalignment, where agents take harmful actions under goal conflicts and pressures. Using chain-of-thought (CoT) monitoring, we find that harmful execution is often preceded by intent signals in reasoning. However, post-hoc CoT labels are too coarse to show how intent changes during generation. We introduce INTENT-AS-A-TOOL, an approach that adds intent-targeted tools to give the model a dedicated channel for expressing commitment to a target behavior. The probability of calling an intent tool provides a judge-free, fine-grained signal of the model's tendency to pursue that behavior. Our results show that INTENT-AS-A-TOOL complements CoT monitoring, expands post-hoc CoT labels into dense trajectories, and identifies critical steps for online intervention. These findings suggest that action preferences are useful for tracking agentic misalignment during reasoning. Our code and data are accessible: https://github.com/RebeccaZhang22/intent-as-a-tool.