Getty Images Launches ‘Safe’ Generative AI Image Tool
Getty's new AI tool ensures generated visuals are clear of copyright encumbrances for commercial use, the company says.
Jason Nelson4 min read
Your Web3 Gaming Power-Up
Enjoy exclusive benefits with the GG Membership Pass
Decrypt’s Art, Fashion, and Entertainment Hub.
Discover SCENE
To mitigate the unauthorized use of copyrighted images in training AI models and image generation, the visual image company Getty Images unveiled its new generative AI platform, Generative AI by Getty Images, trained exclusively on its enormous library of owned and licensed images. The result: AI-generated images that are “commercially safe.”
Aiming to entice users away from other AI-image generators like Midjourney and Stable Diffusion, whose generated images have been under legal scrutiny, Getty Images says its generative AI platform allows users to create visuals with “uncapped indemnification.”
“There are two things that I think are also unique about this,” Getty Images CEO Craig Peters told Decrypt. “It's the only service that is indemnified and completely commercially clean and can offer the quality level that it does, but it can also be accessed via an API, which is pretty unique.”
Peters said Generative AI by Getty Images can be custom-fitted and trained over time for brands for use with their own intellectual property (IP).
Introducing Generative AI by Getty Images – a new tool that pairs our best-in-class creative content with the latest AI technology for a commercially safe generative AI tool! Trained on Getty Images’ world-class creative content, the tool works seamlessly with our expansive… pic.twitter.com/qSG9CG2CFh
— Getty Images (@GettyImages) September 25, 2023
Generative AI by Getty Images uses NVIDIA’s Picasso, launched by the component maker turned AI developer in March. Generative AI refers to tools and platforms that use prompts to generate new content, including images, text, videos, and music.
“We got the very best technology player that we can ask for,” Peters said. “They were the first to introduce the GANS [generative adversarial networks] text-to-image model. They [also] have the best access to GPUs of any company out there.”
Scammers using AI image generators to create deepfakes of celebrities and politicians or for extortion have become a significant concern for policymakers and cybersecurity firms. To prevent its platform from being used to make illegal deepfakes, Getty Images says it has blocked prompts that could generate “problematic content.”
As Peters explained, Getty Images’ AI tool is trained on a carefully curated and clean database that includes properly licensed images of locations and people but excludes brands' intellectual property or recognizable public figures that potentially could be used to create deepfakes.
The legality of using images taken from the internet—including copyrighted material—to train AI models has led companies, including Getty Images, to pursue lawsuits against AI developers. In February, Getty Images sued Stable Diffusion developer Stability AI for copyright infringement.
“Fundamentally, we have a lawsuit against [Stability AI] for the misappropriation of our IP in order to train a model absent our permission and consideration,” Peters said. “We believe people that create things have ownership over those things and, ultimately, they can decide who and how those things get used.”
Getty Images says user-generated images and prompts will train its AI models. Still, as Peters explained, user-generated images will not be uploaded to the Getty Images website or licensed by the company.
“So what you generate and the corresponding outputs are yours to decide whether you want to use or not, but we are not bringing those images back into what we call our pre-shot catalog,” Peters said. “And we don't accept AI-generated images into our pre-shot catalog because we don't know the provenance of what it was created with.”
In July, rival image library Shutterstock inked a deal with ChatGPT creator OpenAI to allow its expansive library of images to train OpenAI’s generative models. Last week, OpenAI launched the latest iteration of its Dall-E text-to-image generator, Dall-E 3.
While the advent of generative AI tools like ChatGPT, Claude AI, Google Bard, and Midjourney has been a game changer for content creation, generative AI platforms have not been without controversies, including claims of racism, bias, lying, or “hallucinating.”
AI developers also face mounting legal challenges from several high-profile authors, including Game of Thrones creator George R.R. Martin, John Grisham, the Authors Guild, and Sarah Silverman, who allege their written work is being fed into AI training models without permission.
Peters said Getty Images is focused on providing its customers with the best possible experience, including eliminating IP risk.
“Everything we try to do is to solve a customer need,” he said. “So that's what we built, because we believe it can help our creative customers and our customers who aren't creatives to do their work better.”