(Bloomberg) — Shutterstock Inc., a marketplace of images and videos long used by content creators, has found a new growth engine: licensing the visual media in its library to train AI models.
Most Read from Bloomberg
Shutterstock’s latest deal was announced Tuesday with Reka AI, a startup that makes large language models — software trained on massive amounts of information that is the backbone of some types of generative artificial intelligence.
The licensing business with AI companies produced $104 million in revenue last year, Chief Executive Officer Paul J. Hennessy said in an interview.
Previously, Shutterstock has said it signed agreements with “anchor customers” worth about $10 million in annual revenue each. Meta Platforms Inc., Alphabet Inc., Amazon.com Inc. and Apple Inc. are among customers for the training data, Reuters has reported. OpenAI inked a six-year deal with Shutterstock in 2023.
“It was critical for us to partner with Shutterstock given its comprehensive source of legally licensable content with best-in-class, human-created metadata,” Reka CEO Dani Yogatama said in a statement.
Massive amounts of data are needed for AI models that power popular tools such as chatbots, which respond to a user’s request by generating text, video or audio. But AI companies are under increasing scrutiny over their use of copyrighted materials in this process. Midjourney, OpenAI and Stability AI built image-generating models with datasets that pulled content from across the internet, a practice that has spurred lawsuits from a number of artists.
Demand for this data has opened up a new opportunity for Shutterstock, whose traditional business of licensing media to advertising firms and creative artists has slowed down in recent years.
Many of the companies that licensed data from Shutterstock “already have much of the world’s content in their own war chest,” but wanted to have images that were legally obtained and contained good-quality descriptions, which assists in the training process, Hennessy said.
Other companies have looked to make money from their information by licensing it in a similar manner to AI companies to build and improve their large language models. Reddit Inc. is being paid roughly $60 million annually by Google for its data and recently signed a separate agreement with OpenAI. Adobe Inc., Shutterstock’s main competitor in its traditional business line, is using its image library to build the company’s AI model, Firefly. Getty Images Holdings Inc., another stock library, has an agreement with AI video startup Runway.
Shutterstock has 771 million images, 54 million videos, 4 million audio tracks and 1 million 3D figures, it said in February. That collection was bolstered by acquisitions of other media repositories in recent years such as Pond5, which focused on videos, and Giphy, which UK regulators forced Meta to divest.
“We were pleasantly surprised,” that Pond5’s video assets ended up being so valuable for AI training, Hennessy said.
Terms of the Reka AI deal weren’t disclosed. Shutterstock projects 2024 revenue for the business unit including the AI licensing will be $138 million, about the same as last year. Hennessy said changing contract structures account for the lack of growth rather than a shortage of potential AI customers. By 2027, Shutterstock expects the segment to produce sales of $250 million.
Most Read from Bloomberg Businessweek
©2024 Bloomberg L.P.
EMEA Tribune is not involved in this news article, it is taken from our partners and or from the News Agencies. Copyright and Credit go to the News Agencies, email news@emeatribune.com Follow our WhatsApp verified Channel
