((Automated translation by Reuters, please see disclaimer https://bit.ly/rtrsauto)) by Katie Paul
A “responsible” AI image generation startup, backed by stock photo provider Getty Images GETY.N, has raised $24 million in a Series A round, including from major advertising agency Publicis Groupe PUBP.PA, she said on Wednesday.
Israel-based BRIA, which licenses more than a billion images from stock providers for its system, said in a press release that it would use the cash injection to expand globally. on a global scale and to strengthen its capabilities in generating text from videos.
The investment, while modest compared to others in the hot field of AI, makes BRIA an early test case for how licensing deals can offer a way forward in the conflicts that make rage over the use of copyrighted content to train generative AI models.
Music labels, artists and other content owners have indicated that licensing is a more acceptable approach to generative AI, after major AI companies built their systems using freely scraped content on the Internet.
BRIA said its licensing approach eliminates the risk of trademarks and other sensitive elements appearing in the images it generates, attracting clients from the world of advertising, marketing and media, including Publicis.
“There is a way to do this that is both responsible and commercial,” Chief Executive Yair Adato said in an interview with Reuters.
Getty, which has supported BRIA since 2022, has played a significant role in litigation over training data.
Last year, Getty sued Stability AI, a leading AI image generation company, for using its photos without a license, which it said was a “brazen” violation of its intellectual property “on a staggering scale”
Last year, it also partnered with chipmaker Nvidia NVDA.O to create an image generation service trained solely on images from its own library.
Competitors Adobe ADBE.O and Shutterstock SSTK.N also offer AI image generators, while all three photography companies license their images to Nvidia for its own service.
BRIA powered its model with visuals from 18 image providers, including Getty, Alamy, SuperStock and Envato. Clients are charged for access to the platform and revenue is shared with photography companies.
Each time BRIA generates an image, it displays the original licensed images that provided data for that result and counts the share of each of them.
Mr. Adato likens this approach to a Spotify model for the AI era, which pays content owners for their role in implementing the technology.
Controlled training also helps prevent technology from being used to create misleading images that could tarnish brands or influence elections.
BRIA’s model, for example, is not capable of generating an image of US President Joe Biden because it was never taught what he looked like, Adato said.