Why Getty Banned AI Images (For Now)

From Plagiarism Today:

Yesterday, Getty Images and iStock have announced that they are following in the footsteps of other art sites, including NewGrounds and Inkblot, in banning artwork generated by artificial intelligence (AI) from their service.

According to Getty’s announcement, “There are open questions with respect to the copyright of outputs from these models….” They further add that the move is to protect their customers from potential legal issues down the road.

Getty noted that this ban does not prevent the submission of 3D renders nor does it prevent the use of any digital editing tool, such as Photoshop or Illustrator.

It’s easy to look at this as another example of a battle line being draw with regards to AI-generated art. However, it’s a bit more complicated than that.

As divisive as AI-generated art is, the Getty ban is very different from the ones we discussed earlier this month. The reason is that Getty works different from sites like NewGrounds and their lack of comfort with AI stems almost entirely from legal concerns, not quantity or quality concerns.

To that end, it’s worth looking at why Getty, most likely, chose to ban AI images and how that could impact AI art moving forward.

Why Getty is Different

Sites like Newgrounds and Inkblot accept user-submitted content and host it on their servers. Because of this, they enjoy a great deal of protection under the Digital Millennium Copyright act (DMCA) should some of that content turn out to be infringing.

In short, under the DMCA, as long as they retain a DMCA agent and remove allegedly infringing content when properly notified, they will be protected under the law. 

However, Getty is in a different position. First off, they have editorial control over the images and then license those images to customers. This greatly increases their legal exposure as they could face repercussions both from directly from artists and their customers.

After all, part of Getty’s service they provide is that they promise customers that they hold the rights to the images they are licensing and, because of that, customers won’t face legal consequences for using Getty-licensed works.

But, with AI, that becomes a problem. All AI systems are trained on earlier works, and how much of those earlier works make into the images that it outputs depends entirely on the AI and the prompt it was given.

The fear is that the AI could “generate” art that is so similar to some of the work it was trained on that it could represent a copyright infringement. However, there’s simply no way for the human creator of the AI work, let alone Getty, to know if that’s a possibility with a given image.

As such, Getty has decided to avoid the issue altogether for now, banning AI art from its service.

Not Just Copyright

Though copyright gets the lion’s share of attention when it comes to legal issues AI art faces, there are other potential problems too.

For example, AI can be used to create images of real people in very unreal situations. We saw this recently with an AI app that swapped the face of actors in porn films with other people, creating the illusion that a different person was in the footage.

Though the AI that performed that was shuttered, there’s no reason other tools couldn’t do the exact same thing.

However, it isn’t just pornography. AI can generate images of individuals in a variety of compromising situations that never happened such as getting arrested, supporting/speaking hate speech and so forth.

All this raises issues of defamation. This is especially difficult for Getty, as many of their customers are news agencies, seeking photographs to accompany their reporting.

Getty is in the business of licensing images that are free of legal issues. In short, they’re selling legal certainty. AI images are the antithesis of that, generating not just countless images, but an equal number of legal unknowns that Getty is not prepared to take a chance on at this time.

Link to the rest at Plagiarism Today

The OP continues, noting that at least one other image-posting/licensing site has prohibited AI generated images.

PG suggests that litigation concerning copyright and ai images is almost certain to happen at some point in the not-too-distant future. If he had to guess about an early plaintiff, he would name Walt Disney, which has made and continues to make a huge amount of money from its cartoon characters.

If anyone sees/hears about an ai copyright case, PG would appreciate an alert through the Contact PG link at the top of the blog.

9 thoughts on “Why Getty Banned AI Images (For Now)”

  1. If anyone is interested in seeing first-hand what these current text-to-image tools do, you can try the basic, Open Source Stable Diffusion by stability.ai for free online at Hugging Face (an online space for AI R&D).

    There are a lot of sources online about how to craft prompts to get what you’re looking for (e.g. this one).

    Stable Diffusion is being built into GIMP, Blender, Photoshop, etc., and can run on your machine (* depending on your graphics card) without requiring “the cloud”. So if you use those tools look out for those plugins.

    If he had to guess about an early plaintiff, he would name Walt Disney,…

    See the section Kermit Anything in How to Write an Awesome Stable Diffusion Prompt

    Warning: playing around with these can be a time sink!

  2. Alternatively, Getty may be positioning itself to have the high ground when it starts suing to “protect” their vast collection of IP. They’re actually the top of my list of people that will act to shut these applications down.

    • They’ve better be prepared to loose. The Google indexing case is a big precedent aside from the fact tbat tbe”AI” art engine output is nothing if not *transformative*.

      I’m thinking they’re just trying to deligitimize their future competition to protect stock image revenue.

      • The questions are: Who is invested in the company? How deep are their pockets? How much of it are they willing to spend?

        You can stop a competitor by running them out of money for legal fees. Or at least make it not worth the money to keep fighting, no matter how deep the competitor’s pockets are. (See Visual J++.)

    • They’re actually the top of my list of people that will act to shut these applications down.

      They may act, but they will fail. Anyone remember how hard the big entertainment companies worked to shut down the ability of new technology to write copies of files it read? No read/write ability because they feared people would use it to copy their stuff.

Comments are closed.