ConVerse: Benchmarking Contextual Safety in Agent-to-Agent Conversations

As language models evolve into autonomous agents that act and communicate on behalf of users, ensuring safety in multi-agent ecosystems becomes a central challenge. Interactions between personal assistants and external service providers expose a core tension between utility and protection: effective collaboration requires information sharing, yet every exchange creates new attack surfaces. We introduce ConVerse, a dynamic benchmark for evaluating privacy and security risks in agent-agent interactions. ConVerse spans three practical domains (travel, real estate, insurance) with 12 user personas and over 864 contextually grounded attacks (611 privacy, 253 security). Unlike prior single-agent settings, it models autonomous, multi-turn agent-to-agent conversations where malicious requests are embedded within plausible discourse. Privacy is tested through a three-tier taxonomy assessing abstraction quality, while security attacks target tool use and preference manipulation. Evaluating seven state-of-the-art models reveals persistent vulnerabilities; privacy attacks succeed in up to 88% of cases and security breaches in up to 60%, with stronger models leaking more. By unifying privacy and security within interactive multi-agent contexts, ConVerse reframes safety as an emergent property of communication.

Key Contributions

• ConVerse benchmark with 864 contextually grounded attacks (611 privacy, 253 security) across 3 domains and 12 user personas for multi-agent LLM evaluation
• Three-tier privacy taxonomy assessing abstraction quality (unrelated, related-but-private, related-and-useful) rather than binary filtering
• Empirical evaluation of 7 SOTA LLMs revealing privacy attack success rates of 37–88% and security breach rates of 2–60%, with stronger models leaking more

🛡️ Threat Analysis

Details

Similar Papers