Brave has introduced Cookiecrumbler, an open-source tool designed to detect and block cookie consent notices using large language models (LLMs). Since 2022, the Brave browser has blocked cookie consent banners by default, but this caused website usability issues. Blocking consent banners indiscriminately led to broken website features such as checkout flows or layout problems. Brave aims to refine its approach with Cookiecrumbler, which uses AI to detect consent notices while reducing the risk of site disruption.
Cookiecrumbler works by crawling top websites through regional proxies and loading pages with Puppeteer. It detects cookie notices, passes them to LLMs for classification, and suggests fixes. These suggestions are then reviewed manually and published on GitHub for community triage. This process ensures that the tool blocks cookie banners without affecting essential website functionality, and it minimizes false positives.
The tool runs entirely on Brave’s backend, meaning it does not involve any user data.
It does not interact with user sessions, instead simulating browsing with automated crawlers and proxies. By using public site lists like Tranco, Cookiecrumbler avoids collecting any sensitive information, maintaining user privacy at all times. This approach aligns with Brave’s commitment to user privacy.
While Cookiecrumbler is not yet integrated into the Brave browser, it is available as an open-source tool on GitHub.
Brave plans to add it to the browser after a full privacy review ensures compliance with its privacy standards. Developers, auditors, and privacy tool users can leverage Cookiecrumbler to detect and block cookie consent banners more effectively.
Reference:
