Technology
Scraper teams up with Cara app founder to build artist tracking tool
Following a series of data scrapes on the artist portfolio platform Cara, the individual behind the largest scrape agreed to collaborate with its founder on an open-source detection tool.
The short version
- The artist-focused portfolio app Cara experienced three major data scrapes in August 2026, exposing millions of image links, personal metadata, and portfolio files.
- A student scraper using the handle Heft expressed regret after witnessing artists' reactions, deleted his 12-terabyte dataset, and joined Cara's Discord to help address security vulnerabilities.
- Cara founder Jingna Zhang and Heft are now building an open-source tool named Lantern, designed to notify artists if their visual work is detected in public AI datasets.
- Zhang raised over $100,000 via GoFundMe to explore legal avenues against unauthorized harvesting while acknowledging that fully preventing web scraping remains technically improbable.
Key facts
- Beginning August 13, 2026, the artist platform Cara was targeted by three separate data scrapes, spiking its server costs and causing some users to delete their portfolios.[Wired]
- The initial scraper, operating under the pseudonyms Heft and MandarinDawnPoppy994, compiled a 12-terabyte archive of approximately 12 million Cara images and posted about it on Reddit.[Wired]
- Subsequent scraping efforts transferred roughly 8.5 million Cara links and metadata to Hugging Face, while a third incident uploaded 123,000 images and user bios to Academic Torrents.[Wired]
- Hugging Face stated it would issue a notice to remove personal metadata but declined to remove URLs pointing to images published on Cara.[Wired]
- Cara founder Jingna Zhang launched a GoFundMe campaign to cover legal fees related to data harvesting, raising more than $100,000 toward a $120,000 goal.[Wired]
- Heft apologized, removed the original dataset, and partnered with Zhang to develop Lantern, an open-source tool that scans public AI image datasets using one-way image fingerprints to alert creators.[Wired]
What remains uncertain
- It remains unverified whether major commercial AI developers utilize scraped datasets uploaded by independent users to platforms like Hugging Face.[Wired]
- The efficacy of legal remedies against data scraping under current copyright and cybersecurity frameworks is clear, as formal legal protections remain unsettled.[Wired]