When something is new and different, the first thing we should do is throw it all on a raging bonfire. That is what Getty Images decided to do last week when it banned and removed all AI-generated art from its marketplace. The image above depicts artwork burning while patrons look on at a Getty Museum-like gallery. It was created with Midjourney, so it cannot go up on Getty Images.
“Effective immediately, Getty Images will cease to accept all submissions created using Al generative models (e.g., Stable Diffusion, Dall-E 2, MidJourney, etc.) and prior submissions utilizing such models will be removed,” according to Voicebot.ai.
The reason? “There are open questions with respect to the copyright of outputs from these models and there are unaddressed rights issues with respect to the underlying imagery and metadata used to train these models,” says Getty. Oh, it’s to protect their customers and artists. There just might be another force at play here.
The Legal Shield
Does the move really protect anyone? Getty points out you can still use Photoshop. Anyone familiar with photoshop knows that you can take any image and then begin manipulating it. In fact, this is a deliberate act. Getty Images’ end user license agreement (EULA) addresses the issue of using someone else’s work.
3.e. No False Representation of Authorship. You may not falsely represent that you are the original creator of a work that is made up largely of licensed content. For instance, you cannot create artwork based solely on licensed content and claim that you are the author.
Note the modifiers “largely” and “licensed content.” Who determines how largely is defined? Why is this only applied to AI-generated artwork? Getty must have artists occasionally using other people’s licensed artwork inadvertently and dressing it up in Photoshop. Licensed images show up frequently on websites with supposedly free images. How are they policing that, and why a blanket ban on AI-generated art alone?
It is fair to say that no one knows precisely what precursor works are involved in AI-generated artwork. There is little doubt that this question will make its way through the court system before long. There is a lot of art that has been digitized, and a lot of it looks similar to something else.
Neal Stephenson, author of Snow Crash and Crytonomicon, said on a recent podcast that his co-founder at a new venture called Lamina1 is working on an algorithm to identify the source artwork that makes up elements generated by text-to-image generators. This topic is going to get a lot more attention very soon.
The Prompt As Content
The basic principle here rests on the age-old question, “What is art?” and the newer question, “Is everything derivative?” If everything is derivative, and an asset is deemed to be art, how do you assign ownership rights of the original content? Of course, there is the legal test about whether the new artwork is transformative. If it is, there is no copyright infringement. This is a subjective test by nature, but it is what we have.
Another key question will be whether the courts and the art industry deem the prompt as content. The key contribution of most AI-artists is the prompt. That expresses the idea and guides the AI system to generate the art. Jason Allen, the winner of one of the prizes at the Colorado State Fair this year, likens his AI tool of choice, Midjourney, to a paint brush.
Ann Spalter agrees. “[AI] is not going to make everyone an artist just like the camera didn’t make everyone an artist, but we all use the camera,” she told me at the Synthedia Conference. Anne also discussed how she uses prompts and how they are a core part of the creative process. It takes knowledge and skill to create a great prompt that will give you the type of image you are hoping for. Some might say, writing prompts for text-to-image generators is an artform itself.
Circle the Wagons
Tommy Mintz, associate professor of art at CUNY Kingsborough College and a professional photographer, said the situation reminds him of the reaction to photography and then to point-and-click photography. Photography was initially rejected by many in the fine art establishment in its first phase, and then the point-and-click version was assailed by professional photographers of that first phase.
Getty Images is clearly part of the art establishment. They may have a plan to sort through the legal challenges and eventually bring AI-generated artwork onto its platform. More likely than not, this is just the typical first step of blocking the innovation until market forces make it untenable to resist any longer. Getty Images may be interested in protecting artists but is definitely interested in protecting its business model which survives on high price points and active legal enforcement. That seems like the primary motivation behind this preemptive move.
With that said, it’s their choice what services to offer. The question now is who will become the Getty Images of AI art? The opportunity is open for the taking.