In a report from Bloomberg via MSN.com, Getty Images is set to debut its own AI image generator in an attempt to generate content free of copyright concerns. These concerns have grown rapidly since a federal court’s ruling found that AI-generated images could not be copyrighted.
The company will use its content library to help generate new images, which likely has given it a great deal of data to train its model. Early this year, Getty sued Stability AI, the creator of the Stable Diffusion AI image generator for using its images without permission.
But the new AI image generator wasn’t solely created by Getty. The company partnered with chipmaking giant NVIDIA to develop the model which was trained on Getty’s image library. As others have reported, by doing this, Getty is likely to avoid legal troubles that other AI services have dealt with due to training data.
An example of this is a recent lawsuit from authors in the United States against OpenAI. This is due to claims that ChatGPT was trained on data that continued data from their copyrighted work. The lawsuit is recent and will likely take some time to resolve.
So by using only data from their content library, Getty is removing a potential issue earlier in its AI journey in comparison to other companies. Another guardrail Getty Images is adding is not allowing users to incorporate trademarked material or assets into the generator.
This would likely reduce the risks of deepfakes and other images that are high risks from being created using the Getty Image generator. But there is flexibility for businesses who wish to use Getty’s product.
For example, the company will allow companies and customers to add their proprietary data or branding to the generator. But, this data won’t be added back into Getty’s libraries, both securing the customer’s IP rights and protecting the company from liability issues.
AI generator images are currently a hot market. With the likes of DALL-E, MidJourney, Stable Diffusion, and others, AI image generation has become both an art for some and an essential business tool for others.
This move came almost a year after Shutterstock’s venture into AI. Finally, the company also announced that images will receive Getty’s usual license to use the content as well as indemnification against suits. Getty will also compensate artists and contributors whose work was used to train the AI model.
So it seems the company is trying to find a balance when it comes to data privacy, content creation, and AI.