Five years after the release of OpenAI's Dall-E and the ensuing controversy over AI-generated images, stock image providers such as Shutterstock have found licensing agreements to be a way to remain viable in the competitive AI market. The deals reflect a broader trend in which content companies — from news publishers to platforms like Reddit — are monetizing their data archives as AI developers seek training corpora, amid ongoing legal uncertainty over whether scraping publicly available content constitutes fair use.

At the outset of the generative AI era, Shutterstock and other image providers, including Getty Images, were forced to respond as the industry underwent a transition in which many AI models were trained on publicly available data scraped from across the web. For Getty Images, the response was to sue image model provider Stability AI, while also creating its own generative AI product, Generative AI by Getty Images. That lawsuit remains a closely watched test case for how copyright law applies to AI model training.

Shutterstock, meanwhile, opted to sign multiple data licensing agreements with OpenAI, Meta and others. The company has also established a contributor fund to compensate artists whose work is included in licensed datasets, an effort to address creator concerns that have accompanied the rise of generative imaging tools.

Daniel Mandell, senior vice president of data licensing and AI at Shutterstock, said the stock image provider views licensing agreements positively.

"We are thinking about data licensing in that two-prong approach, which is model training and then for inference purposes," Mandell said on the Targeting AI podcast from AI Business.

He added that while partnerships might lead to some competition from the AI technology vendors Shutterstock partners with, many users are not just looking for generated images.

"I don't know if it's necessarily replacing the value of stock content," Mandell said, adding that users want content that performs best.

"Customers are moving more toward performance-based content than just having 55,000 pictures of cardinals and nests," he said, referring to bird photos.

"What generative AI is doing is creating a hybrid approach where it gives users more of an ability to create different versions that may be relevant to different customer types," Mandell continued. "I don't think that's something you could have done in the past."