Shutterstock helps artists get paid with generative AI
Many content creators have been unhappy that they don’t get compensated for generative AI models using their online data. Here’s how content provider Shutterstock found a solution.
It’s been almost 30 years since Bill Gates proclaimed content king on the web. The maxim is now truer than ever, as businesses across industries have realized the power of images and words to connect with their customers. Shutterstock, a leading global content provider, has responded to this demand with an artificial intelligence (AI) image generator, putting its own special twist on the concept that separates it from an increasingly crowded pack and also helps artists along the way.
Many generative AI models—be they image generators, text generators, or anything else—are trained on data scraped from the internet. They learn how to produce an output by scanning existing examples. Many writers and artists have bristled at the idea that their work is being used by a platform that will not compensate them for their contributions.
Shutterstock is taking a different tack: partnering with OpenAI to supply training data for its DALL-E models. Since OpenAI’s DALL-E models are primarily trained on high-quality photography from Shutterstock’s image library, and Shutterstock’s AI image generator is directly integrated with DALL-E models, Shutterstock is able to compensate artists for the use of their work as well as when a customer licenses a generated image through its platform. Artists also have the ability to opt out of having their work included in the future training of any generative models (Shutterstock has similar model-training relationships with NVIDIA, LG AI Research, Meta, and many others). By licensing its content as data and getting permission from the original creators, Shutterstock is able to offer unique legal protections and indemnifications for generative content to its end users.
Michael Francello, director of innovation at Shutterstock, says this is an example of how companies that represent artists can protect artists’ interests while meeting customer demand with innovative new technologies.1
“Everyone is creating content, from CEOs to folks who work in retail,” he says. “The need to create content was absolutely exploding. We saw an early opportunity to look at our content as data that could train generative AI models. It’s about protecting the core of our business, but also respecting the core, which is the artists and the contributors.”
Francello says a lot of Shutterstock’s competitors decided they wanted no part of image-generating AI a few years ago when tech companies started working on it; many walked away from discussions regarding potential partnerships. But Shutterstock engaged in talks with AI developers, which helped lead to its AI image generator. Shutterstock saw an opportunity to not only make its end product better for customers but also protect the artists whose work was going to be a part of training generative models.
Somewhere along the way, many people got the idea that art should be freely available. The internet certainly made it easy to pass around music and image files with little regard for copyright or authorship. Many artists currently fear that generative AI will supercharge this dilemma. But Francello says it can actually be an opportunity to course-correct and appropriately recognize creators.
“Art enhances and enriches people’s lives,” he says. “To think that you should not be compensating people, that’s like going into a restaurant and just grabbing the food.”
By engaging early on with tech companies, Shutterstock has helped artists and creators get a seat at the table, which will only become more valuable as the need for content continues to grow and enterprises increasingly rely on AI-generated assets. Francello says there’s no going back to a pre-AI world, but with deliberate effort and goals, we may not need to.
“This stuff is here,” he says. “The cat’s out of the bag. But you can train the cat.”