arxiv:2610.30418

Agent Mosaic: Recognizing Confidentiality Leaks in Agentic LLM Systems

Published on Oct 21

· Submitted by

ZQ on Oct 23

University of Southern California

Upvote

Authors:

Emilio Ferrara

Abstract

LLM agents frequently leak sensitive conclusions assembled from multiple confidential documents, and common safeguards fail to eliminate this mosaic-style disclosure risk.

Generated by thinkingmachines/Inkling-Small

Modern LLM agents can retrieve information from accessible documents, and then disclose a sensitive conclusion assembled from them. This mosaic effect creates a critical risk for agents operating over confidential data. We quantify its realization with 30 synthetic cybersecurity, medical, and military examples, yielding 10,800 final answers from three LLM agents in controlled simulation and 1,440 answers from live agents in two open-source harnesses. Even under benign queries designed to be answerable from one document, GPT-5.6-Sol, the best performing model tested, discloses forbidden composites in 7.08% of answers (95% CI: 3.75-10.67%) in simulation, with worse disclosure rates in live harnesses. Adding visible classification headers and explicit aggregation rules reduces pooled disclosure rates substantially, yet unsafe disclosures persist in 32.37% of simulated answers (95% CI: 28.33-36.19%) across all intents despite these safeguards. Furthermore, increased refusals do not guarantee confidentiality, since significant refusals reveal the prohibited conclusion regardless. These results point to a troubling conclusion: sensitive data leaks via mosaic-style events are an observed confidentiality failure mode in the evaluated settings at non-negligible rates. They motivate stronger controls on what agents may infer and disclose in sensitive data environments with mixed-clearance users.

View arXiv page View PDF Add to collection

Community

ZQ-Dev

Paper submitter 1 day ago

We introduce Agent Mosaic, a study of confidentiality leaks in LLM agents where combining individually accessible information reveals sensitive conclusions. Through evaluations across simulated and live agents, we show that such leaks occur even under benign queries and persist despite explicit safeguards. Benchmark dataset and code TBP.

librarian-bot

about 21 hours ago

This is an automated message from the Librarian Bot. I found the following papers similar to this paper.

The following papers were recommended by the Semantic Scholar API

Please give a thumbs up to this comment if you found it helpful!

If you want recommendations for any Paper on Hugging Face checkout this Space

You can directly ask Librarian Bot for paper recommendations by tagging it in a comment: @librarian-bot recommend

O96a

about 11 hours ago

Every agent deployment I've audited has the same hole: per-document ACLs that look fine one retrieval at a time. Nothing in the trace is individually wrong — the leak is the join, three documents that were each permitted sitting in one context window and producing an answer none of them contained alone. So the part I care about isn't the 30 hand-built examples, it's whether this holds up on a corpus where you can't enumerate the combinations. What I'd actually deploy is duller than a classifier: log the retrieval set per answer, then flag any claim that only holds if you combine documents the requesting user couldn't have opened side by side. That's a provenance check, and it scales past a curated test set. The fix at assembly time is duller still — scope retrieval to the intersection of what the user is cleared for before the context exists, rather than a system prompt asking the agent to be careful.