If you use Adobe Generative Fill to process one of your photos, do you own the copyright? If you add, remove or expand your image, is it still your image?
What happens when you use Photoshop Generative Fill?
You probably have heard of Generative Fill by now. Adobe has incorporated Firefly (AI-generated art app) into Photoshop (beta). This allows you to add, remove, and extend your image by typing in text prompts. Firefly is trained on Adobe Stock’s hundreds of millions professional-grade licensed, high-resolution images. Adobe says that this makes sure that Firefly won’t generate content based on other people’s work, brands, or intellectual property.
However, we also know that as part of Adobe’s Content Credentials feature, AI images made in Photoshop will be encoded with an invisible digital signature indicating whether it’s human-made or the product of AI.
Above: The photo on the left is a photo I created. The photo on the right has been largely expanded using Adobe Generative Fill. Do I still own the rights to this photo?
But if you add, remove, or expand your photo using Generative Fill, do you still own your image? Can you still copyright it?
If someone takes your image and expands the size to create a much larger image, can they claim this as recontextualized, new art? Can they copyright that?
What does the United States Copyright Office say about this?
In 2018, The Copyright Office denied a copyright claim because the work “contained no human authorship.” They further ruled that it was made “without any creative contribution from a human actor.”
Flash forward to 2023. The Copyright Office approved a graphic novel. They concluded that a “human-authored text combined with images generated by the AI service Midjourney constituted a copyrightable work.”
Don’t get too excited just yet. They also ruled that the individual images themselves could not be protected by copyright.
So far, the Copyright Office is consistent in that copyrights will only be issued if the image is the product of human creativity. In fact, they state that the term “author” excludes “non-humans.”
You can read more about their rulings about artificial intelligence in the Federal Register.
What else is the Copyright Office doing?
The Copyright Office is currently examining “copyright law and policy issues raised by artificial intelligence (AI) technology, including the scope of copyright in works generated using AI tools and the use of copyrighted materials in AI training. After convening public listening sessions in the first half of 2023 to gather information about current technologies and their impact, the Office will publish a notice of inquiry in the Federal Register.”
Until then, it appears like it’s the Wild West. Artificial intelligence and the programs that use it are moving faster than the Copyright Office can keep up.
Photoshop (beta) – no commercial use
Please note that as of this writing, Generative Fill is not for commercial use while in the Photoshop (beta).
Opinions?
What’s your opinion? If you use Generative Fill, do you feel that you own your image? Is there a point in which you feel someone no longer owns their image? If someone expands the size of your image much larger, can you still claim ownership?
I think it depends entirely on what’s generated and the extent is modified by yourself afterwards. If I take something from stock images to composite into my scene, I should be able to copyright the result as a derivative work. Same with something like a sky replacement…I fail to see the difference between a replaced sky from a stock image or a replaced sky generated in AI. I suppose it’s a 50%+1 thing. As long as >50% of the image is mine I feel comfortable saying it’s my image, therefore it should be copyrightable. Hopefully the regulations around this will… Read more »
Yes, and I hope that it’s relatively sensible. On the face of it, if you cut and paste something or do sky replacements without AI, you would think it’s the same. But with AI-generated (non-human) art becoming such a huge part of a photograph, it seems like we are now entering the domain of work that is not “human-made” as defined by the U.S. Copyright Office. And that seems to be where things could get murky.
My thinking is if I took the photo it is mine to do with what I please.
The copyright issue is irrelevant unless I am trying to sell it. Even then it is MY image and if I add AI in editing it still came from me. What if I use AI to reduce noise? Same thing. It shouldn’t really matter as long as the image is acceptable to the person who created it in the first place.
Yes, Dave, when isn’t it considered generative? Isn’t Content Aware a form of generative?
I would imagine both are regarded as generative. There is a difference, however.
Content Aware pulls information from your image only.
Generative Fill pulls from information your image and Adobe Stock from Firefly and is therefore also informed by images other than your own.
You can do what you please with your photo (which is the whole idea with the sort of technology I am discussing in the article). And that is likely correct about the copyright.
I should point out that we are discussing AI-generated art becoming an enormous part of the photograph, which is rather different from employing AI to get rid of noise. Generating mountains or a giraffe is completely different from fixing hot pixels in your photo.
One can only hope. It’s the “human made” component that seems to be up in the air right now. Thanks again for your comments, greatly appreciated.
You are making the decisions. The Ai is only your assistant that is either regenerating pixels that are already in your photo (what’s the difference between this and a plugin that makes your photo look painted?), or it’s going to the stock house to find other Human Made images (for now) that might work with yours. Those other authors should be compensated for the use of their images, but only as Stock. I have a photo where I made a night photo in a garden still life looks like it was underwater. I (not an AI) sourced one photo of… Read more »
Hey Ken, Is there’s way to fine tune Adobe GF with for example 1,000 Flikr images of say “travel posters” that Adobe Stock might not have. (Or maybe Adobe has borrowed everything already) So someday…“make a travel poster that is 50% AS data and 50% using my Flikr folder.” Sorry, newbie.
No. Currently, GF is powered by Firefly although it obviously can also pull elements from your own image as well. Firefly is informed by Adobe Stock, which Adobe feels is more ethical than pulling from many photos from elsewhere, whoch might have copyright issues.
On those rare occasions when I change a sky to remove lens flares, I use content “I” generated under the same conditions (preferably in the same location around the same time) to do it. Period. That way, there’s no question of copyright ownership or my ability to register my work with the copyright office. It’s up to me to previsualize what I want in a photograph, then choose or create conditions to render it in camera. I may be a bit of a purist, but I try to render what’s in front of me, or what I light-painted there myself.… Read more »
Can Adobe see the images we work on while using AI or neural filters?
Since they are not allowing certain types of images to be made, how will that be enforced?