Singaporean photographer Zhang Jingna is leading a campaign to protect artists’ work from being harvested by artificial intelligence (AI) scrapers without consent. Zhang, a fashion and fine art photographer based in the United States, founded the social platform Cara to provide a safe space for artists to share their work with controls against unauthorized AI data collection.

Cara, which has approximately 1.5 million registered users, employs “NoAI” tags designed to signal to web crawlers and bots that images should not be used for training AI models. Despite these measures, the platform was targeted three times in August by human scrapers who extracted images from Cara’s servers to supply generative AI developers. These actions took place without the consent of Cara’s community and have caused significant concern about artists’ rights and data misuse.

The first scraper, identified by the pseudonym “Heft,” copied over 12 million images from Cara. Heft initially believed the images were free to use for AI development but later apologized and removed the content after recognizing the harm caused. Nevertheless, his actions inspired a second scraper, who used generative AI tools to harvest images and shared them on the AI development platform Hugging Face. Despite multiple takedown requests, this data remains accessible.

Zhang disclosed the continued scraping incidents via social media in mid-August and revealed the third recent breach on August 22. Cara, launched in 2023 and still in beta, is volunteer-run and primarily funded by Zhang, incurring additional operational expenses due to the scraping activity. She has since initiated a fundraising campaign that has raised about US$70,000 to support potential legal actions against the scrapers.

Zhang highlighted the emotional toll on herself and the artistic community, stating that even with options to opt out, their work is being exploited and that these attacks feel personal. She described scraper behavior as not merely technical violations but acts that cause distress to creators relying on online exposure for commissions.

Defenders of AI scraping argue that publicly posted images are accessible for training datasets, with some commentators suggesting that if creators do not want their work used, they should refrain from posting it online. However, the legality of such scraping remains ambiguous. In the United States, scraping publicly available data is not explicitly illegal, though it conflicts with privacy principles, according to a recent legal review. Meanwhile, Singapore is currently conducting a public consultation, launched on August 26, to determine whether copyrighted works may be used for AI training, with submissions accepted until October 22.

Following his shift in perspective, Heft has joined forces with Zhang to combat unauthorized scraping and even developed a tool to help Cara users detect if their images have been harvested during subsequent attacks. Zhang expressed concern that advances in AI tools make scraping easier and more difficult to prevent, creating ongoing challenges for artists.

Zhang advocates for stronger legal protections for artists who share their work online, emphasizing the urgent need for clear regulations to safeguard creative content against exploitation by AI developers and unauthorized scrapers.