Stability AI Boss Admits to Using ‘Billions’ of Images Without Consent

Dangers of AI

Yesterday at a Senate Judiciary Committee hearing on artificial intelligence (AI) and copyright, Stability AI head of public policy Ben Brooks admitted to using “billions” of images without asking the copyright owner’s consent to train the AI image generator Stable Diffusion.

While this is now a well-established fact, with other AI image bosses like Midjourney’s David Holz also admitting to using troves of copyrighted works without consent, Brooks was made to feel uncomfortable during the grilling with artist representative Karla Ortiz only too happy to stick the boot in.

Senator Mazie Hirono put it to Brooks that he didn’t pay to use any of the content used to train the AI image generator.

“If that image data is out on the internet and robots.txt says it can be subject to aggravated data collection and if it’s not subject to an opt-out request in our upcoming models then certainly we will use those images,” responds a visibly uncomfortable Brooks.

Earlier in the hearing, Brooks had dodged a question from Senator Chris Crooks on whether Stability AI pays for its data.

When pressed further on the matter, with Senator Hirono saying “So basically you don’t pay for your data” there is another awkward moment.

“For the base, the initial training of these models with those billions of images, there is no arrangement in place,” says Brooks.

When Senator Hirono turns to Ortiz, a concept artist and illustrator who works on major movies, she points out what so many photographers and artists have been saying about AI image generators.

“I have never been asked, I have never been credited, I have never been compensated one penny,” says Ortiz.

“And that’s for the use of almost the entirety of my work both personal and commercial.”

Ortiz went on to say that she would never consent for her work to train an AI and stressed that companies like Stability AI do not give her a choice.

Every photographer in the world who has ever posted a photo online will be in the same boat as Ortiz. Stability AI, which operates Stable Diffusion, uses the LAION dataset a “publically available” data pool that can be searched. The website Have I Been Trained allows copyright holders to see if their work was included in the data model.