Search papers, labs, and topics across Lattice.
The paper introduces CI-Work, a Contextual Integrity benchmark designed to evaluate privacy risks in enterprise LLM agents that retrieve and utilize internal context. Evaluations of state-of-the-art models on CI-Work reveal significant privacy violations (15.8%-50.9% violation rates, up to 26.7% leakage), and a concerning trade-off where higher task utility correlates with increased privacy leakage. The authors argue that addressing these vulnerabilities requires a shift towards context-centric architectures rather than simply scaling model size or reasoning depth.
Enterprise LLM agents leak sensitive information in up to 50% of interactions, and surprisingly, performing better at tasks makes the problem *worse*.
Enterprise LLM agents can dramatically improve workplace productivity, but their core capability, retrieving and using internal context to act on a user's behalf, also creates new risks for sensitive information leakage. We introduce CI-Work, a Contextual Integrity (CI)-grounded benchmark that simulates enterprise workflows across five information-flow directions and evaluates whether agents can convey essential content while withholding sensitive context in dense retrieval settings. Our evaluation of frontier models reveals that privacy failures are prevalent (violation rates range from 15.8%-50.9%, with leakage reaching up to 26.7%) and uncovers a counterintuitive trade-off critical for industrial deployment: higher task utility often correlates with increased privacy violations. Moreover, the massive scale of enterprise data and potential user behavior further amplify this vulnerability. Simply increasing model size or reasoning depth fails to address the problem. We conclude that safeguarding enterprise workflows requires a paradigm shift, moving beyond model-centric scaling toward context-centric architectures.
