From manual rules to AI automation: scaling cookie popup protection at DuckDuckGo
summary
DuckDuckGo automates opt-out by letting CMPs load and programmatically clicking reject flows, combining tracker blocking and cosmetic hiding when needed. A JSON-rule library called auto consent encodes site-specific steps; thousands of site-specific rules generate and update automatically, with crawls of ~10,000 sites per region and coverage ~90% on top sites. AI is used only for lightweight classification of cookie UI text, plus scripts and regexes to extract buttons, and a validation pipeline reduces false positives below ~1%, making the system practical at scale.
AI-written summary · about
chapters (7)
- 0:00Introduction to talk
- 1:07Problem with cookie popups
- 5:06DuckDuckGo’s popup protection
- 11:09Scaling challenges and long tail
- 13:17AI for rule generation
- 16:55Future directions and open source
- 36:03Q&A session
related talks
20mHow to defend your sites from AI bots
43mNext in Privacy Innovation: AI Agents that Make Governance Scalable
36mData privacy in an AI-driven world
01h 19mEffective Transparency and User Controls: Are We Close to a Breakthrough?
16mWhy Consent Fails in Practice: Lessons from Web Measurement Studies
24mBots & Humans: Maintaining AI agent and automation in an AI-bot web
20mWho Left Open the Cookie Jar?
16mWebsites' Global Privacy Control Compliance at Scale and over Time