Contextual Integrity in LLMs via Reasoning and Reinforcement Learning
This addresses the challenge of preventing privacy leakage and inappropriate information sharing in autonomous AI agents, representing an incremental improvement with a novel hybrid method.
The paper tackled the problem of ensuring contextual integrity in LLMs by prompting them to reason about appropriate information disclosure and developing a reinforcement learning framework, resulting in substantial reductions in inappropriate information disclosure while maintaining task performance across multiple models and transfer to human-annotated benchmarks.
As the era of autonomous agents making decisions on behalf of users unfolds, ensuring contextual integrity (CI) -- what is the appropriate information to share while carrying out a certain task -- becomes a central question to the field. We posit that CI demands a form of reasoning where the agent needs to reason about the context in which it is operating. To test this, we first prompt LLMs to reason explicitly about CI when deciding what information to disclose. We then extend this approach by developing a reinforcement learning (RL) framework that further instills in models the reasoning necessary to achieve CI. Using a synthetic, automatically created, dataset of only $\sim700$ examples but with diverse contexts and information disclosure norms, we show that our method substantially reduces inappropriate information disclosure while maintaining task performance across multiple model sizes and families. Importantly, improvements transfer from this synthetic dataset to established CI benchmarks such as PrivacyLens that has human annotations and evaluates privacy leakage of AI assistants in actions and tool calls.