Cara, an image-sharing platform built for artists opposed to AI training on their work, was hit by three separate large-scale data scrapes in August 2026, exposing millions of images and user details and prompting its founder to launch a legal fundraising campaign that has raised more than $100,000.
The first incident began on August 13, when a Reddit user posting under the name MandarinDawnPoppy994 shared a 12-terabyte archive of approximately 12 million images from Cara — effectively its entire publicly available library — to the subreddit r/DefendingAIArt. The user described it as “a fun project” that cost less than $10. Cara’s founder, photographer Jingna Zhang, says the platform learned of the scrape after users flagged the post, where the individual was “gloating and looking for other people to join him.”
Two further scrapes followed. A second actor, identified as “CaptiveDreamer,” pulled roughly 8.5 million links along with metadata including usernames, titles, and tags, uploading them to AI developer platform Hugging Face. Hugging Face said it would ask the user to remove personal metadata but declined to act on the URLs, stating that no copies of the artworks were hosted on its platform. A third scrape on August 22 collected 123,000 images, text posts, and user bios containing personal information, which were shared on a site called Academic Torrents.
Zhang has described the attacks as “targeted and very hurtful,” noting that laws have not kept pace with protections against data harvesting of this kind. She is separately involved in two ongoing class-action lawsuits — one against Stability AI, Midjourney, and others, and a second against Google — alleging their image generators were trained on copyrighted work.
In an unexpected development, MandarinDawnPoppy994 later expressed regret and agreed to collaborate with Zhang on an open-source tool aimed at protecting artists. Meanwhile, Zhang launched a GoFundMe with a $120,000 goal to pursue legal strategies under cyber and copyright law. As of late August 2026, more than $100,000 had been raised, and Cara said it is actively seeking additional legal assistance.
Cara, maintained by Zhang and a small volunteer team since early 2023, has attracted around 1.5 million artists. The platform already offers protective features including Glaze, a tool designed to mask image styles from scrapers, but Zhang acknowledges that preventing scrapes entirely is “nearly impossible.”
Source: WIRED