How Artists Are Fighting Back Against AI Art That Copies Their Works

Artificial intelligence (AI) is shocking the art world with its ability to produce creative images with the click of a mouse, and now artists are looking for ways to protect their works. 

Researchers at the University of Chicago have created a software called Glaze to keep AI models from learning a particular artist's style. It's part of a growing effort to prevent AI art from including the work of human artists without attribution. 

"Artists need to be aware that AI algorithms are able to recognize patterns and replicate them in a highly efficient way," Michael Osterrieder, the CEO of vAIsual, a company that uses algorithms to generate synthetic stock media, told Lifewire in an email interview. "This means that digital collections of AI artwork can, and are, being used to train AI  to recognize patterns, so it can integrate the styles and compositions that look similar to the original artwork it learned from. However, artists need to understand that without their consent to be included in AI training, it is not fair use."

Protection From AI-Generated Art

Glaze is intended to hide images so that AI models incorrectly learn the unique features that define an artist's style, preventing later efforts to generate art which leads to AI plagiarism.

"Artists really need this tool; the emotional impact and financial impact of this technology on them is really quite real," computer science professor Ben Zhao said in a Glaze news release. "We talked to teachers who were seeing students drop out of their class because they thought there was no hope for the industry, and professional artists who are seeing their style ripped off left and right."

The researchers came up with the idea of using AI to work for artists instead of against them. They turned to 'Style transfer' algorithms, which use an existing image and recreate it in a particular mode without changing the content.

Glaze uses this process on original art, identifying the specific features that change when the image is transformed into another style. Then the software returns to the source and perturbs those features just enough to fool art-mimicking AI models while leaving the original image almost unchanged to the human eye.

"We're letting the model teach us which portions of an image pertain the most to style, and then we're using that information to come back to attack the model and mislead it into recognizing a different style from what the art actually uses," Zhao said.

Legal Protection From AI-Generated Art

Dvir Ben Aroya, the CEO of Spike, said to protect their artwork from AI, artists can add digital watermarks or signatures to their work, making it harder for an AI system to replicate the artwork without the watermark. They can also use low-resolution images of their work online or create different versions of their artwork that are not easily reproduced by an AI system.

"In addition, artists can also copyright their work, which provides legal protection and allows them to take legal action against anyone who infringes on their copyright,” Aroya added. “However, it can be difficult to prove that an AI system has copied an artist's work, as AI-generated works may not be exact replicas of the original.”

Lawyers for artists are arguing for creative human protection in court cases against Stable AI and Midjourney. The cases “should hopefully set a powerful precedent that will protect artists against their work being used without their consent,” Osterrieder said. 

Osterrieder’s company is working with content creators to package their information into datasets sold to the companies training the AI. The company is also working on creating unique  AI models that "celebrate the distinct style of living artists and help to amplify and scale their creativity.”

The concern about AI-generated art isn’t just limited to visual arts. David Ciccarelli, the CEO of Voices.com, an online voice talent marketplace, said in an email that AI is ripping off voice artists. “Voice-overs artists and companies looking to hire a voice-over talent to train AI voices should take proactive steps to establish mutually agreeable firm terms of use, similar to a code of conduct or a licensing agreement,” Ciccarelli added. “Consent and transparency will be critical to exploring generative AI in an ethical way.”