CMU School of Drama


Monday, August 24, 2026

Artists Built A Site To Escape AI. Scrapers Are Coming For It Anyway

www.forbes.com: Cara, an image sharing social network built specifically for artists who do not consent to have their work used to train generative AI models, suffered a third scraping attack in the past ten days, with copyrighted images and metadata taken from the site without permission appearing on the open source data network Academic Torrents on August 22. This follows previous scrapes of the dataset posted on Reddit (since taken down by the original poster) and the AI model clearinghouse Hugging Face.

4 comments:

Lex McDonald said...

This was both a very interesting and alarming read. For me personally, I despise the use of most generative AI and see its data collection actually harming and undermining the consent of the artists who create. Being at CMU has also opened my eyes to just how many people use AI, and as an artist, it is gravely concerning to see how my work can be taken and fed into a giant machine, and unless I have enough money to take action, there is nothing I can reasonably do. What was also alarming was the fact that these seemed like normal people scraping Cara, not major corporations, which is terrifying to think because, unlike corporations, people cannot be regulated or punished as harshly, and there is much more likelihood of them getting away with it and companies benefiting from artists' work. Seeing the amount of AI used also concerns me, as it suggests there is real demand for it, and that same demand could be taken from human people who may rely on the internet for their income, such as online art creators.

Cate Baltodano said...

I think the concept of Cara is great–the message that real artists are against ai and actively not consenting to having their art be fed to ai is a very important one. It is saddening that the article is about scrapers and how they’re actively stealing from these talented artists. I found it appalling when the scrapers were defending themselves. I think the argument that ‘we NEED to feed the models so that they can develop better’ is absurd. The special thing about art is that it is unique per person and needs the time to develop, by having years of experience be fed to a machine to produce slop takes away what makes art art. Despite what many might claim about their intentions with ‘training’ ai, there are always the people using it for horrible, and sometimes borderline illegal purposes. As amazing as the idea of Cara is, unless there is a direct way to stop people from screenshotting or scraping the site’s art the ai issue is likely to not go away.

Caitlyn W said...

This article really highlights the growing problem of AI stealing the jobs and livelihood of millions of people. This article and the people behind "Cara" have a deep care for art and its comforting to know that there are people out there trying to build safe spaces for artists to form a community built on authentic talent and creativity. The way AI and the people behind it become more aggressive consistently attacking "Cara" using childish reasons such as "an eye for an eye" shows how unplanned and dangerous it is that these attacks are happening. The people on AI's side and the mechanisms we are feeding our info to, to train them do not seem to understand the worldwide problems it causes. People will continue to lose trust in art and culture now due to AI's mission to replicate it and take over. Hopefully "Cara" continues to be a safe space for artists that is provided the funding to protect the space and give artists an opportunity to speak to other artists not robots.

Octavio Sutton said...

I think this article brings up a really interesting topic for artists. What does implied consent mean legally and how do we avoid being used in AI training and programs. I agree with the part of the article that mentions it is impossible to be fully protected from scraping data from the internet. However, I think it's important to keep spaces that allow for people to share their art without having to consent to it being used in training models or datasets. I would like to learn more about how sites like Cara would hold up in a legal sense and what they can do to protect themselves from scraping. What I think is most important to take away from this article is that we are barely keeping up with the advancements in technology. Because of that, its super important to understand what you are consenting to online now and anticipating how things might change in the future.