ARTISTS ARE TAKING THINGS INTO THEIR OWN HANDS TO PROTECT THEIR WORK FROM GENERATIVE AI

The oil painting depicts a woman standing on a podium, her arm aloft as she grasps a laurel crown in her hand.

- 

A scarlet cloak drapes across her chest as she stares at the viewer. To the naked eye, the painting looks like a normal piece in an online portfolio. But the version of the painting uploaded online belies a hidden defense system — a tool called Glaze that masks the artist’s style and cloaks the art from use by generative AI.

As image-generating AI continues to evolve, artists have increasingly fought against what they see as an existential threat to their craft on multiple fronts: through lawsuits, in public statements calling for regulations and now, with programs aimed at protecting their art from being scraped and emulated without their permission.

Created by researchers from the University of Chicago, Glaze and a second program, Nightshade, essentially poison the well of art uploaded online in an attempt to scramble what AI sees. Where Glaze subtly changes an image so AI perceives it as a different art style, Nightshade is a more “offensive” tool that attempts to confuse an AI training model about what is in an image. The idea, university researchers said, is to provide a technical solution to stop the “malicious” use of AI models while also protecting creators.

The team working on Glaze and Nightshade noted that their tools serve as a safeguard in a space that lacks regulation rather than as a comprehensive solution. Where lawsuits and government regulations might force tech giants like Microsoft or OpenAI to change how they operate, smaller AI players outside of the United States might not follow suit, researchers said. In those instances, the AI scrambling tools will still be useful.