AI Image Firm Accused of Copyright Infringement

The makers of an artificial intelligence tool for "creating" art are being sued by a photo licensing company. The case could set a precedent for how copyright law works with modern technology.

The case involves Stability AI, the company behind a "deep learning, text-to-image model." It's designed mainly to create images based on a text description provided by the user.

As well as being a neat trick in itself, the idea of the model is to develop computer learning. That means that rather than humans providing a set of rules to follow, the computer model figures out rules itself. Imagery is a particularly useful field for this research as spotting patterns in pictures and recognizing images is traditionally one of the few activities where humans remain much faster and more accurate than computers.

Unlicensed Use

In this case, Stability AI was "trained" using hundreds of thousands of images found online. It used the images and attached descriptions to find connections, to the point that it doesn't just recognize the content of images, but also the style.

The dispute is about whether that use was legal. One of the sources of the photos was Getty Images, which licenses photographs for commercial use. It says that Stability AI used its images and associated metadata without permission for the software training.

Getty believes it's a straightforward case as it has previously licensed its images for AI research by other companies. It says that failing to get (or even seek) a similar license, Stability AI has clearly breached its copyright on the images and data. (Source: gettyimages.com)

At the time of writing, Stability AI had yet to comment on the legal action, saying it had not formally received documents detailing the complaint. (Source: bbc.co.uk)

Fair Use Claim

The company has spoken out about similar claims of infringement and says it is protected by fair use principles. Other defenders of the technology have argued that the use of copyrighted images for "training" is no different to a human artist studying other artists and copying their style.

If the case goes to trial, a court may have to decide whether that argument still holds up with automated technology, and how to set a threshold between imitation and infringement. It may also have to decide whether Getty having previously licensed images for AI research means it has the right to say who can and can't use its images in that way.

What's Your Opinion?

Which side should win this case? Should AI companies need to get permission before using other people's content for training their models? Should copyright only cover direct copies of material?