Getty Images to debut its own artificial intelligence image generator

Photography giant Getty Images Holdings Inc. is releasing an artificial intelligence tool that will generate images from the company’s vast content library — an attempt to create AI content free of the copyright and ownership concerns that have plagued the technology so far. 

Getty, which holds the rights to millions of photographs, earlier sued Stability AI, the company that popularized the image generator Stable Diffusion, for using images without permission. Getty’s new product, developed with chipmaker Nvidia Inc., will be trained Getty’s own data and will endeavor to sidestep thorny legal issues, in part, by limiting what images will power the generator.

The new tool will tap Getty’s bank of creative images, but not its news photo collection, part of an effort to prevent the creation of deepfakes, Chief Executive Officer Craig Peters said. The new image generator won’t allow users to incorporate trademarked material or assets they don’t own — so there’s no way to create something like the viral Pope Francis wearing a Balenciaga puffer coat image, Peters said.

In order to cater to businesses looking to create ads and other content, Getty will allow customers to add their own proprietary data or branding. Content generated through the product, which will create images based on text prompts, won’t be added back into Getty’s own libraries.

The AI-generated images will receive Getty’s usual license to use the content, as well as indemnification against suits. The company also said it plans to compensate artists and contributors whose work was used to train the AI model. 

The growing popularity of text-to-image AI — such as OpenAI’s Dall-E, Stable Diffusion and the Midjourney service — have raised questions about whether those tools benefit from the work of artists, photographers and designers without getting their permission or compensating them. In addition to Getty’s suit against Stability AI, various artists have also sued services including Stable Diffusion and Midjourney. 

Getty will continue to invest in its legal action against Stability AI in the US and the UK, Peters said.

Uncertainty around the technology means that companies that want to use AI software to create new images for uses like ad campaigns or social media posts worry they may open themselves up to legal jeopardy and fines, said Peters, who noted that Getty’s customers have repeatedly raised this concern. 

“There are real risks here,” he said. “Customers want to use generative AI but do not want to run into this sort of minefield of, ‘We don’t even know if we own this thing.’”

Earlier this month, Microsoft said it will defend buyers of its artificial intelligence products from copyright infringement lawsuits and pay related fines or settlements.

The new Getty service shows that AI companies that say they can’t develop the technology while respecting intellectual property rights are not being truthful, Peters said. “It fundamentally undercuts one of the arguments of those that put these generative models out there without compensating,” he said. –BLOOMBERG