Task graph approach improves privacy in multi agent communication workflows
Judging Before Propagation: Task-Graph Contextual Integrity for Multi-Agent Privacy
Cryptography and SecurityArtificial Intelligence
Summary
Many computer systems now use multiple agents that talk to each other and remember past conversations. This can cause privacy problems because information shared might be safe when given once, but not if it spreads or combines with other data later. The authors created a way to track and control how information flows through these agents, looking at the whole process instead of single messages. Their method reduces privacy leaks a lot without hurting how well the system completes tasks.
What this means in practice
- •For enterprise software developers: Prevent privacy breaches in enterprise applications that use multiple AI agents sharing information across workflows by enforcing broader disclosure controls.
- •For cloud service architects: Design cloud services managing coordinated agent interactions with better guarantees about information sharing compliant with privacy policies throughout a task workflow.
Authors
Hengyu Wu, Blaise Delattre, Yang Cao
Abstract
Multi-agent LLM systems increasingly rely on persistent memory and inter-agent communication. These capabilities create privacy risks that extend beyond individual interactions: a release appropriate in isolation may become inappropriate when combined with information previously disclosed to a recipient or propagated to downstream agents along the planned workflow. Existing Contextual Integrity (CI) mechanisms for agentic systems typically rely on local judgments that assess each disclosure in isolation, with or without regard to the recipient's history, but do not account for how information propagates downstream. However, compliance at the current boundary does not guarantee appropriateness across the workflow. We introduce Task-Graph Contextual Integrity (TG-CI), to our knowledge the first task-graph-native formalization of Contextual Integrity for multi-agent systems. TG-CI represents a workflow as a Multi-Agent Task Graph (MATG) and extends CI from atomic transfers to stateful, graph-level information flows by jointly considering an item's intended propagation scope and recipients' evolving disclosure states. We operationalize TG-CI through propagation-scoped authorization under two complementary policies: Full-Scope, which requires admissibility across the complete downstream scope, and Auto-Region, which confines each item to its maximal admissible region using lightweight frontier gates. We further introduce MAPLE-Bench, a benchmark of 144 cases across four scenarios that jointly vary MATG topology, disclosure history, compromised-agent placement, and item-level privacy policies. In our primary experimental configuration, TG-CI reduces the mean unauthorized disclosure count by 85.8-94.1% relative to the Local-History CI baseline, while retaining 93.6-96.1% of the baseline's task accuracy, with the two enforcement policies providing complementary privacy--cost--utility trade-offs.