Shutterstock and Getty Images, two of the most popular stock photo sites have announced a partnership with OpenAI and BRIA, respectively to allow users to create custom pictures on their platforms after shunning AI-generated content last month over copyright concerns.
Shutterstock will be using OpenAI’s DALL-E software while Getty Images will be using BRIA’s in-house image generation API. These AI image generation models take a simple phrase as input from the user and present them with a full image reflecting the phrase.
However, due to the nature of these AI models, copyright claims still remain an issue with these new partnerships flirting with the limits. Text-to-image models are trained on large datasets made of photos and illustrations or other art resources found on stock image sites. These photos are often scraped without explicit consent meaning with an accurate enough phrase, the AI can mimic an artist’s style, putting the final image in a copyright grey zone.
Shutterstock states that it licensed its images and relevant metadata to OpenAI in 2021 to train its DALL-E model in 2021, however, there is data from other sources in the training set as well. The AI will be integrated into Shutterstock over the coming few months allowing users to generate and customise their own images, There’s no launch date announced for the feature yet.
As for creators on the platform, they’ll be compensated if their photos or art ends up in DALL-E’s training set with royalties to be paid for intellectual property.
The Getty-BRIA partnership doesn’t come with a launch date either, with the companies simply announcing their common goal to “allow creatives to transform images to their specific needs using intuitive AI tools on Getty Images’ platform”.
There’s also no word on how the partnership will compensate creators if their assets end up being used to train BRIA’s AI or how they’re going to tackle potential copyright infringements.