AI scrapers killed Gentoo's bug tracker. The open-source project had already banned AI.

AI training scrapers forced Gentoo Linux to shut down its Bugzilla bug tracker, making the service unusable for the volunteer developers and users who depend on it. Gentoo Council member Michał Górny, who took the service offline, had spent months documenting how automated crawlers were overwhelming the project's servers. The tracker is now in a recovery phase following a database migration, with mail notifications only partially restored 1.

The shutdown lands on a project that explicitly rejects AI. Gentoo's Bugzilla front page carries the instruction: "Please do not include any work from AI in your bug reports, per Gentoo's AI policy" 1. The Gentoo Council voted unanimously in April 2024 to ban contributions created with AI tools, citing copyright, quality, and ethical concerns 2. Now AI scrapers have taken down the infrastructure those bug reports depend on.

A data pipeline bots already had, and ignored

The detail that separates this from a generic server-overload story sits on Gentoo's own bot policy page. Gentoo already built dedicated infrastructure for automated crawlers: static data dumps of every public bug, refreshed every two hours, available in compressed form, with explicit rules about which URLs bots may access 3. The policy was designed to let scrapers get everything they need without touching the live database.

AI scrapers ignored all of it. Górny wrote in an April 2026 blog post that LLM crawlers were "firing Bugzilla search after search, report after report," overloading the server and database while real users could not file or find bug reports 4. He compared the cumulative effect to a sustained distributed denial-of-service attack on independent infrastructure 4. Gentoo had already locked Bugzilla search behind a login wall to keep database-heavy queries away from unauthenticated crawlers. When that was not enough, Górny took the tracker down entirely.

How Gentoo got here

The escalation from policy stance to infrastructure collapse spans more than two years:

• February 2024: Górny proposed banning AI-generated contributions on the gentoo-dev mailing list 2.
• April 2024: The Gentoo Council voted unanimously to adopt the ban, covering code, documentation, bug reports, and commit messages 2.
• Prior to shutdown: Gentoo locked Bugzilla search behind a login requirement after crawler queries degraded database performance 5.
• April 2026: Górny documented the scraper crisis, writing that LLM crawlers were ignoring Gentoo's bot rules and overloading the database 4.
• This week: Górny took Bugzilla offline. The service is now recovering, with a database migration completed and mail notifications partially restored 1.

Who pays for the harvest

Gentoo is not the only project absorbing this cost. A commenter on the Hacker News thread reported that Hedgewars, an open-source game, locked down its own Bugzilla for the same reason 5. LWN.net appended a note to its 2024 Gentoo article stating: "The LWN site is currently under high scraper load, so comment display has been suppressed for anonymous users" 2. The outlet that reported on AI scraper damage is itself running under AI scraper load.

HN commenters speculated that the worst offenders were likely Chinese AI training projects operating from Southeast Asian IP ranges, though this remains unconfirmed 5. Another commenter observed that identifiable crawlers from major AI companies were "normally fairly well behaved," and that the heaviest traffic came from bots disguising themselves as ordinary Chrome browsers 5.

The structural issue is cost externalization. A company training a model pays nothing per request to scrape a volunteer-run bug tracker. The project absorbing those requests pays in server load, developer time, and degraded service for the humans it was built to serve. Platforms like GitHub can deploy enterprise bot protection and absorb the bandwidth. Gentoo, maintained by volunteers, cannot match that spending.

Gentoo's bot policy, data dumps, and login walls were each reasonable countermeasures. Each was overwhelmed. If the solution is heavier gatekeeping across open-source infrastructure, the access cost falls on the same communities the infrastructure was built for, not on the companies sending the bots.