Using OpenAI, Shutterstock will begin selling stock photos created by AI.
Will the image industry be destroyed by AI image generators? Many people are asking this question now that text-to-image AI algorithms have become more prevalent in recent years. The response from established players in the market, however, is “no” — not if we can first begin selling AI-generated material.
Today, the online stock photo company Shutterstock announced an expanded relationship with OpenAI, which will result in the direct integration of AI Lab’s DALL-E 2 text-to-image model into Shutterstock “in the coming months.” Additionally, Shutterstock will create a “Contributor Fund” to pay authors when they sell their work to the company in order to train text-to-image AI models. This comes after these methods received harsh criticism from artists whose work was taken from the internet without their permission. Notably, Shutterstock forbids the selling of art produced by AI on its platform if it wasn’t made via the DALL-E integration.
Third-party AI art is forbidden at Shutterstock, and it provides its own DALL-E integration.
Paul Hennessy, the CEO of Shutterstock, stated in a press release that “the media for expressing creativity is always growing and expanding. We acknowledge that it is our duty to embrace this change and make sure that the generative technology that fuels innovation is based on moral principles.
“We are thrilled that Shutterstock can provide its clients DALL-E photos as one of the first delivery through our API,” said Sam Altman, CEO of OpenAI. “We look forward to further collaborations where artificial intelligence becomes an intrinsic part of creative workflows of artists.”
Shutterstock and OpenAI have worked together before on projects like these. Shutterstock began selling images and metadata to OpenAI in 2021 to assist in the creation of DALL-E (according to OpenAI’s Altman, this data is “essential to training DALL-E”). The cooperation has now fully circled with the addition of OpenAI’s text-to-image AI, and the output of DALL-E will now face competition from the same people whose work was used to train it.
Artificial intelligence (AI) art generators may produce a wide range of images, however they are taught using data downloaded from the internet, frequently without the authors’ permission.
If Shutterstock’s images were as crucial to the development of DALL-E as Altman claims, the platform’s contributors may reasonably be upset that their work has been used to eliminate jobs for others. This is also the reason Shutterstock is introducing its Contributor Fund, which will be used to compensate authors, photographers, and designers when their work is used by businesses like OpenAI to create generative AI models.
It’s a big move and the first time a platform owner has taken such a big step to pay creators in this fashion, but it also highlights how thorny the legal and moral issues are with this new technology.
Although it appears lawful to scrape or purchase data to train AI art generators (protected by fair use), several experts are concerned about potential difficulties and issues. Because it is unable to copyright the output from these systems, Getty Images, for instance, has prohibited the sale of AI art on its site due to worries that this may cause licencing problems for users.