Brave brings AI muscle to block annoying cookie consent pop-ups

Brave has open-sourced its Cookiecrumbler tool, an AI-powered system that automatically detects annoying cookie consent pop-ups. Developers and privacy advocates can now join the fight against changing banners.

Brave, the privacy-focused web browser, wants to fight cookie consent banners with AI and the community.

A new tool, Cookiecrumbler, automatically detects cookie consent notices on websites using open-source large language models (LLMs). Brave announced this tool last year.

It is not yet user-ready and is under active development. Brave expects that it will help deliver an automated, large-scale process to detect new and changing banners.

No to cookies – no need for consent banners

Two years ago, Brave started automatically blocking “annoying and privacy-harming cookie consent banners.” However, this feature is limited. It is based on “EasyList Cookie,” a filter list maintained by a team of volunteers. EasyList filters are widely used to block ads, trackers, and annoyances online.

The main challenge is that blocking cookie consent notices sometimes breaks essential website functionality, such as checkout flows and layout. This limits adblock list maintainers as they rely on as few general rules as possible.

Each site implements cookie consent notices differently, and generic rules don’t cover all cases. Brave said it spent months cleaning community-maintained blocking lists and removing parts that cause problems.

“We’ve encountered many issues (broken scrolling, blank pages) when a cookie consent notice block is applied indiscriminately.”

Brave blocks third-party tracking scripts and pixels by default to avoid situations where websites don’t respect user choices, even when selecting to reject all cookies.

“Many consent systems still track people, even when users reject all cookies. These notices are especially useless in Brave, since we already block third-party tracking scripts and pixels,” Brave said.

How does Cookiecrumbler work?

The Cookiecrumbler tool automatically tackles this problem with surgical precision. It’s a site-specific pop-up blocking solution. AI helps automate detection, including non-English cookie consent notices.

“LLM-based automation frees up human reviewers to do the more nuanced work of actually blocking the notice,” Brave explains. “It’s cheap, as every crawl costs on the order of cents.”

The basic idea is that Cookiecrumbler suggests what to filter out, and human maintainers will either confirm or tweak these suggestions before they go live. The quality of the automated process is expected to improve over time.

Cookiecrumbler works by loading a website on a headless browser (Puppeteer), using proxies for the desired region. Then it identifies and gathers HTML elements of the cookie consent notices. The LLM then checks the notices and suggests fixes before sending the response back. Currently, the tool answers an automated crawler, which publishes results on a GitHub repository.

Brave is considering building Cookiecrumbler into users’ browsers, enabling smart cookie notice detection. For now, it's under privacy review.

Cookiecrumbler is still a work in progress. Since its introduction last year, Brave has “made significant progress in lowering false positives,” added support for multiple languages, and ensured coverage across multiple geographical vantage points.