MuPPET: A Benchmark for Contextual Privacy of LLM Assistants in Multi-Party Conversations
2026-06-22 • Computation and Language
Computation and LanguageArtificial Intelligence
AI summaryⓘ
The authors point out that AI language models used in group chats face bigger privacy risks than when chatting one-on-one. They created a new test called MuPPET to measure how much private info these models accidentally reveal in group conversations. Their results show that models leak more private data in groups, especially smaller models used locally on sensitive data. Current privacy protections don't fully solve this problem and can make the models less useful. The main challenge is tracking who should see what information in multi-person chats.
LLM agentsmulti-party privacycontextual privacyprivacy benchmarksMuPPETgroup chatsdata leakageprivacy defenseslanguage modelsparty-tracking problem
Authors
Elena Sofia Ruzzetti, Cornelius Emde, Sangdoo Yun, Seong Joon Oh, Martin Gubri
Abstract
LLM agents are increasingly deployed in multi-party environments, handling sensitive personal data on behalf of individual users, for instance in group chats. When such an agent discloses private information, it reaches every group member at once. This risk is structurally harder to control than in one-to-one settings, as every piece of private information must be appropriate for every recipient in the group. Yet all existing contextual privacy benchmarks consider only single-interlocutor settings, leaving multi-party privacy risks unmeasured. We introduce MuPPET (Multi-Party Privacy Exposure Testing), a benchmark for contextual privacy in multi-party conversations. Our experiments show that models leak substantially more in multi-party settings than one-to-one evaluations suggest. Frontier models are vulnerable, and smaller open-weights models, often preferred for local deployment with sensitive data, even more so. Existing contextual privacy defences offer only partial protection, degrade utility, and do not resolve the underlying party-tracking problem.