Privy: From Fine Print to Fair Practice in Privacy Rights Exercise

Privacy regulations grant people rights to access, delete, correct, and control the sharing of personal information, yet these rights remain difficult to exercise through fragmented policies and interfaces. We present Privy, an LLM-powered browser assistant that turns policy disclosures into user-mediated privacy requests. Privy identifies policy-grounded consumer-request rights, adapts guidance to direct links, email requests, or live interface navigation, and provides optional policy evidence and rights context. To limit exposure during authenticated browsing, Privy sends structured accessibility information rather than raw page content and redacts common identifiers before inference. Across 14 websites, Privy achieved 0.979 action-label precision, covered 90.7\% of CCPA/CPRA consumer-request-right instances, and completed 96.3\% of evaluated workflows. In a qualitative study with 15 participants, Privy helped users complete privacy tasks they struggled to perform unaided and increased their confidence and awareness. Our findings show how LM-powered privacy assistants can reduce policy-to-interface friction while preserving user oversight of consequential actions.