Polisis: Automated Analysis and Presentation of Privacy Policies Using Deep Learning
summary
Polisis uses a privacy-focused language model trained on 130K privacy policies and a hierarchy of 21 neural classifiers to label policy segments, then answers structured and free-form queries. It assigns Disconnect-style privacy icons with 88.4% accuracy against law-student labels and powers PriBot, which finds a correct answer among its top-3 82% of the time; a Mechanical Turk study finds at least one top-3 answer relevant for 89% of user questions.
AI-written summary · about
chapters (3)
related talks
01h 13mCan automated processing make privacy notice/choice more effective ...
01h 17mPrivacy and New Technologies
43mPanel: Privacy Design Patterns for AI Systems: Threats and Protections
01h 20mUnlocking societal benefits of AI with privacy protective technologies.
16mAutomating Technical Privacy Reviews Using LLMs
28mPrivacy Engineering Networking Break AMA
11mHow Does AI Change Privacy? Towards Helping Practitioners Scope, Motivate, and Conduct
16mPrivacy in the Public Sector: Lessons Learned and Strategies for Success