Gentoo shut its public bug tracker after AI scrapers overwhelmed it

Gentoo, a volunteer-run Linux distribution, took its public bug tracker offline this week after a sustained surge of AI scrapers overwhelmed the project's donated infrastructure. The shutdown, announced by Gentoo developer Michał Górny on Mastodon, makes the project one of the most prominent open-source casualties yet of a string of AI-scraper-driven shutdowns of public infrastructure.

A bug tracker, in this case the Bugzilla instance at bugs.gentoo.org, is the public site where users file, search, and discuss software defects. For a Linux distribution, it is also the central coordination layer between developers and the people running the operating system. Closing it to public access cuts off the workflow that lets a volunteer project survive.

Ars Technica reported in March 2025 that open-source developers were blocking entire countries to keep AI crawlers from saturating their bandwidth. InfoQ documented in February 2026 a series of AI-driven "vibe coding" floods that pushed small projects offline. Gentoo's published policy at bugs.gentoo.org/bots.html is a working example of the kind of bot list other maintainers consult when they draft their own. The post this week is what happens when policy and infrastructure both run out of room.

Discussion on Hacker News gives a clearer picture of the trade-offs than the announcement, because it is where maintainers describe what they have tried and where each option breaks. The cheapest move is bot-aware load balancing through a service like Cloudflare. It works until scrapers learn to spread across IP ranges, and the load balancer slows real users along with the bad ones. IP and user-agent filtering is more precise but produces false positives. Commenters note that scrapers cluster on IPv4 ranges disproportionately associated with South-East Asian IP space, which makes a naive block a regional one. Per-page micro-payments, the older idea floated in the discussion, would shift the cost to the scraper but add friction to legitimate readers, including the small developers and researchers who use public bug trackers as a learning resource. Commenters also flag browser-for-hire evasion, where a scraper routes requests through a real browser running on a paying customer's machine, as already observed from at least one well-known AI operator.

Górny has already written about the underlying problem. In an April 2026 essay on his Gentoo blog titled "The pinnacle of enshittification, or Large Language Models", he argues that the LLM industry's economic model treats the open web as a free input and the people who run that web as a cost the industry does not have to pay. The essay is a maintainer's account of the system he is being asked to subsidize while he debugs it. The volunteer project absorbs the scraper cost in donated hours and donated bandwidth. A commercial scraper is a line item in someone else's quarterly model, billed against a budget that exists because the volunteer project did.

Gentoo's shutdown is concrete, attributable, and recent. Load balancing, IP filtering, and per-page fees are the three main options the Hacker News discussion surfaces, and each has a documented cost. Load balancing slows real users. IP filtering produces false positives that look regional. Per-page fees add friction for legitimate readers. Browser-for-hire evasion is already a working counter to the most sophisticated of these. The access restriction Gentoo picks when it reopens the tracker will be the next concrete data point on which trade-off a project of its size can afford.