Despite significant advances in aligning Large Language Models (LLMs) with safety and ethical standards, adversarial prompts can still be crafted to elicit undesirable responses. Existing mitigation strategies predominantly rely on post-hoc filtering, which introduces substantial latency or computational overhead, and is incompatible with token-level streaming generation. To address this, we introduce Self-Sanitize, a novel LLM-driven mitigation framework inspired by cognitive psychology, which emulates human self-monitor and self-repair behaviors. The framework utilizes a lightweight Self-Monitor module to inspect high-level intentions via representation engineering and a Self-Repair module for in-place correction without separate review dialogues. This design allows for real-time streaming monitoring and seamless repair, with negligible impact on latency and resource utilization. Extensive experiments across three privacy-leakage scenarios and broader safety-sensitive domains (e.g., crime and bias) demonstrate that Self-Sanitize achieves superior mitigation performance with minimal overhead and preserved model utility. Our code is available at the following link: https://anonymous.4open.science/r/LLM_Self_Sanitize-5E8B

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.