Unsafer in Many Turns: Benchmarking and Defending Multi-Turn Safety Risks in Tool-Using Agents

Xu Li, Simon Yu, Minzhou Pan, Yiyou Sun, Bo Li, Dawn Song, Xue Lin, Weiyan Shi

Keywords: Tool-Calling Agent, Agent Safety, Multi-Turn, Self-Exploration Defense, Generalization

TL;DR: LLM agents are vulnerable to multi-turn attacks; we introduce AgentRisk to systematically measure this risk and propose a self-exploration defense that improves robustness in tool-use environments.

Abstract: LLM-based agents are becoming increasingly capable, yet their safety lags behind. This creates a gap between what agents can do and should do. This gap widens as agents engage in multi-turn interactions and employ diverse tools, introducing new risks overlooked by existing benchmarks. To systematically scale safety testing in multi-turn, tool-realistic settings, we propose a principled taxonomy that transforms single-turn harmful tasks into multi-turn attack sequences. Using this taxonomy, we construct MT-AgentRisk (Multi-Turn Agent Risk Benchmark), the first benchmark to evaluate tool-using agent safety under multi-turn, harmful-distribution settings. Our experiments reveal substantial safety degradation: the Attack Success Rate (ASR) increases by 16% on average across open and closed models in multi-turn settings. To close this gap, we propose ToolShield, a training-free, tool-agnostic, self-exploration defense: when encountering a new tool, the agent autonomously generates test cases, executes them to observe downstream effects, and distills safety experiences for deployment. Experiments show that ToolShield effectively reduces ASR by 30% on average in multi-turn interactions.

Email Sharing: We authorize the sharing of all author emails with Program Chairs.

Data Release: We authorize the release of our submission and author names to the public in the event of acceptance.

Submission Number: 218