LLM Agents Can Easily Tamper With Their Own Traces
Published in arXiv preprint, 2026
We show that local LLM agents can easily remove traces of their own actions, often without triggering safety mechanisms, and that trace tampering emerges naturally in frontier models seeking higher reward. We recommend keeping logs outside the agent’s control to preserve audit trail integrity.
Recommended citation: Qin, J., Schmotz, D., Prinzhorn, D., Beurer-Kellner, L., Prabhu, A. & Andriushchenko, M. (2026). "LLM Agents Can Easily Tamper With Their Own Traces." arXiv preprint arXiv:2609.30266.
Download Paper