How New Tech Can Prevent Artists' Work From Being Copied By AI

Artificial intelligence is under fire for appropriating images

  • Increasingly popular AI programs that produce art are drawing criticism from artists who say their work is being appropriated.
  • A new range of software is designed to credit artists when AI uses their work. 
  • Websites like can show if specific images were used to train an AI model.
An artificial intelligence drawing a portrait of a man.

Devrimb / Getty Images

Generative artificial Intelligence (AI) is increasingly used to appropriate artists' work, but new technology could help. 

Nvidia is launching software that helps AI models comply with copyright laws and has announced a new program to share revenue with creators. It's part of a movement to ensure that artists are fairly compensated. The issue is becoming more urgent as generative AI programs grow in popularity. 

"Art will evolve into art and code, and part of the art will be protecting it, but like everything, if the art is viewed as valuable, people will try to steal it," Jake Maymar, the vice president for innovation at The Glimpse Group, told Lifewire in an email interview. "Right now, creating art in the artist's or artists' styles is easy to do. The solution is to make it harder to do this by adding other AI technologies of identification, anti-coping, and security that can evolve as fast as the AI models."

Giving Credit to Artists

Nvidia is collaborating with the stock image service Getty Images to train "responsible" generative text-to-image and text-to-video models. The models will be developed using content from Getty Images' library, Nvidia says, with Getty Images providing royalties to artists on any revenues generated from the models. Instead of scouring the web for images, the models are trained on licensed data which Nvidia says ensures they comply with copyright law.

The problem of art appropriation stems from the fact that new generative AI techniques are making it easier than ever to create images, and computer models are built using existing art.

"Using Midjourney, Stable diffusion, or another AI art generation model, an artist can train the model on a specific art style," Maymar said. "Once the model has enough of the reference artist's style, it can then create unique, similar artwork based on that chosen artist's style. This process can be automated to create hundreds of thousands of art pieces in any artist's style as well as mixing and matching different artists' styles together."

AI image generators aren't really copying the art they're trained on. "It's more like a very sophisticated, massive-scale collage," Ellery Kreloff, the CEO of Artsitree, a site for digital artists, said in an email. 

A History of Borrowing

Using technology to copy artwork is not new, Edward Balassanian, the CEO of Aimi, a generative music platform, pointed out in an email interview. Cameras and tape recorders allowed paintings and music to be recorded, copied, and sold. Digital copying allowed copying to be industrialized. 

"As with any form of technology, the same tools used to create authorized copies enabled bootlegging as well," he said. "This gave rise to the Napsters of the world, where prolific, mass-scale copying threatened to destroy the economics that enabled artists to invest time in being creative. AI is no different. It will be both a tool that enables a new industry by allowing ethical and authorized use of AI technology to accelerate artists' efforts and a tool to plagiarize them."

Someone choosing from a variety of images in virtual reality.

metamorworks / Getty Images

Nvidia isn't the only company to try to prevent intellectual theft or borrowing of images. Maymar pointed out that sites like Have I Been Trained show if your artwork was used to train the model. He said a model could be trained to identify your art style, and when it appears in a commercial work, the system notifies you of where it is and how it is being used, then an AI lawyer could file a lawsuit based on its findings.

"It is likely that for each advancement in training, there will be an equal investment in tools and techniques to determine if models were trained unethically or illegally," Balassanian said.

Kreloff said that because AI art generation is still in its infancy, the art industry is still discovering the long-reaching impacts.

"For digital artists, many worry they will lose commission work for certain commercial projects: album covers or book covers, for example," she added. "However, there will always be people who value human-made pieces: the same way Etsy can exist alongside Amazon."

Was this page helpful?