Getty Images has launched an artificial intelligence tool that turns written prompts into pictures, pairing access for its subscribers with a plan to pay artists whose work helped train the system. The company also says it will protect business customers against potential copyright disputes tied to the product.
A generator trained on Getty’s library
Developed with chip company Nvidia, the tool was trained using Getty’s own image collection. Getty’s archive contains more than 135 million copyrighted images, according to the source article.
Text-to-image systems learn from examples, connecting written descriptions with visual elements such as objects, concepts, and artistic styles. Getty’s approach uses its licensed archive as the training source rather than a broad collection gathered from across the internet.
Getty chief executive Craig Peters said images produced by the tool will belong to the customer who creates them. They will not be added to Getty’s existing content libraries for others to license. Peters said this separation is intended to keep authentic photography distinct from AI-generated imagery in the company’s databases.
Protection for business customers
Getty pledged uncapped indemnification for more than 800,000 users connected to the product. In practical terms, the company says it will take on legal and financial responsibility for its business customers if potential copyright disputes arise from using the generator.
The promise comes amid growing uncertainty over how copyright applies to generative AI. Getty’s product launch followed Microsoft’s announcement that it would provide indemnity coverage for potential copyright claims arising from use of its AI CoPilot services, which bring generative AI into Word, Excel, and PowerPoint.
Getty is also involved in a legal dispute over AI training. The company filed a copyright claim against Stability AI in the UK High Court, alleging that the image-generation company had “unlawfully copied and processed millions of images protected by copyright.”
Artists are promised recurring payments
Getty says artists whose images helped train the system will receive payments on a “recurring basis,” as Peters described it. He said the initial amounts would be small, while arguing that a growing market for generative AI could eventually create material revenue streams.
The payment plan addresses one of the central questions facing AI image generators: how to account for the creative work used to build them. Peters said creatives’ expertise and investment in their content should be rewarded. He also argued that products built on protected, higher-quality image collections can produce better results than systems trained by indiscriminately scraping the web.
That position is part of a wider debate involving commercial tools from OpenAI, Midjourney, and Stability AI. The source article describes models trained on billions of images from places such as social media, ecommerce sites, blogs, and stock image archives.
A different model for commercial AI imagery
Getty’s launch arrived as other image-generation products were expanding. OpenAI had recently updated DALL-E, and Dall-E 3 was integrated with paid versions of ChatGPT, where users could ask the chatbot to generate images and help develop more detailed prompts.
OpenAI had also said artists could ask for their work to be excluded from future versions of its text-to-image models through a removal form on its website. In May, chief executive Sam Altman told the US Congress that “creators deserve control over how their creations are used.”
Getty’s model combines several commitments: training on its own library, recurring payments to contributing artists, customer ownership of generated images, and legal protection for business users. Whether these measures resolve the wider copyright debate remains open, but they show one way an AI image service can try to address the concerns raised by artists and commercial customers.