Getty Releases AI-Image Maker Trained on Company’s Data

Getty Images announced Monday the launch of a proprietary artificial intelligence tool that will generate images from its wide-spanning collection of digital media. The move will allow the media company, which holds ownership rights over millions of news, stock, and archival images, to avoid copyright issues that have plagued artificial intelligence companies.

On Monday, Getty announced that it was joining the emerging market of AI image creators by releasing Generative AI by Getty Images, a software developed in partnership with tech giant Nvidia that was trained using the companies data and limited photographic content. Development of the new tool was started before Getty initiated the Stability lawsuit, according to a statement.

In a statement, Getty Chief Executive Officer Craig Peters called the new AI tool “commercially safe” and said it will not use images from its news collection for the generator, instead only tapping into the company’s vast creative library for the generative model. The move, Peter said, is a part of the company’s aim to avoid distributing deepfakes, a form of altered media used to disseminate misinformation. Getty’s image generator will block users from using assets it doesn’t have reproduction rights over or that are otherwise trademarked.

The Seattle-based company said it plans to compensate contributing artists whose work was used as data to develop the new AI model. Over the last year, numerous artists and digital media creators have sued generative A.I, companies over the use of their images and works in the development and training of AI technology. In the most prominent class action lawsuit, artists alleged that Stability AI, Midjourney, and DeviantArt violated their copyrights in the use of the images. While there has, as yet, been no decision, it seemed likely in July that the judge would dismiss the suit.

In February, Getty sued Stability AI, the tech company that owns Stable Diffusion, a text-to-image generator, accusing it of “brazen infringement” in using more than 12 million photographs from the stock image company’s collection to train its AI and, according to the complaint, in an attempt to “build a competing business.” That lawsuit is ongoing.

Peters lauded the new software as a more “commercially viable” option for content users, who he said are currently more vulnerable to a “minefield” of issues related to AI image makers that were developed on open-sourced unlicensed imagery on the Internet. He added that the tool’s revenue-sharing model in the release also “undercuts” commonplace claims from other technology companies that AI-related models can’t work in favor of artists whose works are used in the software.