AI Clone of AO3 Taken Down in a Day: 2,000 People Sent Takedown Notices

⚡ The Gist in 5 Seconds

• An AI clone of Archive of Our Own (AO3), built via mass scraping, was taken offline in less than a day.
• More than 2,000 people individually sent the site's creator copyright infringement notices.
• The site's name, the owner's name, and the hosting provider are not mentioned in the story — only a complaint counter and the clone's lifespan.

🔍 What Happened

Jason Koebler of 404 Media wrote on Bluesky that a clone of Archive of Our Own, built with AI and mass scraping of fanfiction, ran for about a day before being shut down. The reason: more than 2,000 people individually sent its creator takedown notices. This was a case of "vibe coding": the site was assembled by prompting a language model to write the code, not by a development team.

The mechanics matter more than the arithmetic here. The authors didn't write a single collective petition or wait for a platform's decision — they sent thousands of independent demands. In the scheme Koebler describes, each such letter is addressed personally to the owner: one complaint, one incident, each of which would have to be handled individually. The clone didn't fall to moderation — it fell to the volume of incoming legal mail.

🧠 Our Take

This story isn't about AI "failing." It's about the barrier to entry dropping to zero: assembling a clone of a major archive is now cheaper than ever, and that applies not just to the code but to harvesting other people's texts as well. It looks like the first mass case where a community of authors stopped a scraper not through the courts or via hosting support, but through a coordinated stream of formal letters.

For authors, this is an effective tactic, and our assessment is that it will become a template. The legal force here lies in quantity, not volume: 2,000 individual notices turn a small pirate site into a project with an unmanageable workload for a single person. For platforms, it's bad news: the same trick can easily be turned against a legitimate service, and no one will be able to tell an author's complaint from a bot's.

Who loses? Those who bet on "scrape everything and sort it out later." Who wins? Archives with a working complaint system and those who can quickly mobilize a community. But there's a third, most interesting player — AO3 itself: based on experience, stories like this tend to hurt those who store other people's texts, not those who generate them.

⚠️ What We Don't Know

We don't know the site's name, the owner's name, the hosting provider, or the volume of works scraped. It's unclear whether it shut itself down or was taken offline by its provider, and whether any of the 2,000 letters were actual DMCA demands. A journalist's post is not a court document, and what needs checking here is the primary source: the clone's page and the text of at least one notice.

Prediction: within a quarter, at least one public service will appear that collects takedown notices from authors into a single package and sends them on behalf of the community. Easy to verify by watching for launches of such tools.

Disclaimer: not officially confirmed; single source — a post by a 404 Media journalist.

Source: Jason Koebler, Bluesky