AI generated fill images and copyright concers

[Janeames]

I am working on a composite where I want to use head swaps and thought of using generative fill.  If I sold these finished composites, would I open myself to lawsuits if someone recognized their image.  I am presuming these downloads are from photographs of people in cyberspace, right?

[D Fosse]

If AI-generated, they are presumably not real people - but theoretically, I suppose they could accidentally be close to some real person.

I'd be more concerned with your own copyright if you use AI-generated material extensively. In short, there isn't one, if the AI material is a substantial part of the finished product. Not to mention the ethical aspect of profilting from something you didn't create yourself.

But this all depends on context. It's not black and white.

[jane-e]

If I sold these finished composites, would I open myself to lawsuits if someone recognized their image.

By @Janeames

Here's what Adobe says:

8. Commercial Use

In general, you may use outputs from generative AI features commercially. However, if Adobe designates in the product or elsewhere that a beta version of a generative AI feature cannot be used commercially, then the generated outputs from that beta feature are for personal use only and cannot be used commercially.

https://www.adobe.com/legal/licenses-terms/adobe-gen-ai-user-guidelines.html 

Jane

[c.pfaffenbichler]

I am presuming these downloads are from photographs of people in cyberspace, right?

I think you might be misunderstanding the nature of generative AI. 

It does not »cut and paste« parts of existing images exactly. 

[Janeames]

Ok. If they r not other people's images ie could be my sister or thr person next door and are totally ai generated if this image gets used and gets notoriety is it wholly mine or does firefly or other entity have any legal or financial claim over part of it?



Sent from my T-Mobile 5G Device
Get Outlook for Android<>

[D Fosse]

It depends. You have no copyright on the ai-generated content itself. Nor does anyone else - for copyright to apply, it has to be created by an actual human.

So the question is how significant the ai contribution is. What is the context? Is there human artistic expression as a whole, or is it mainly machine made? That's food for the legal profession.

Actually, it seems pretty similar to sampling in music. You can do it, but you can't go overboard.

[namuwilliams]

"You have no copyright on the ai-generated content itself. Nor does anyone else - for copyright to apply, it has to be created by an actual human."

While this is partly true, it's important to remember that the U.S. Copyright Office has officially stated that you can copyright the AI part of an artwork as long as there was "significant" human input during the creation of the AI part...the problem is, they're still hashing out what "significant" means, in this case.

"Actually, it seems pretty similar to sampling in music. You can do it, but you can't go overboard."

And this is 100% incorrect. You can't just sample something in music as long as you don't "go overboard". Even a 1 second sample needs to be cleared if the music isn't in the public domain (or unless it falls under "fair use", which is a very limited and strict area of use).

The reality is that some AI art can be copyrighted and some can't, currently, and unfortunately the powers that regulate this sort of thing are taking a long time to present clear rules and guidelines. That's probably a good thing, but I know it can be frustrating for those of us already deep into AI creation, and wanting to make use of our work. Half the problem, I think, is that most of the people making the decisions about this industry don't have any real, in-depth experience in the actual creation of AI art, so they need to rely on outside expert opinion, which slows the process down.

[D Fosse]

Well, OK, but basically this comes down to common sense. If you're pushing a button and a result comes back, you haven't created it and copyright doesn't apply. That's the basic a priori premise and a self-evident axiom.

The gray zones come into play when that push-button result is used as an element in a bigger context, and that bigger context is itself a genuine creation. Sometimes, but not always, that will correlate to a simple area-of-total-image ratio.

Ultimately, you will know the answer to this question yourself. If you're a person with integrity, we can just take your word for it. The problem, of course, is that there are people without integrity who are are happy to take shortcuts. So we need to have an outside check.

[jane-e]

...it's important to remember that the U.S. Copyright Office has officially stated that you can copyright the AI part of an artwork as long as there was "significant" human input during the creation of the AI part...the problem is, they're still hashing out what "significant" means, in this case.

By @namuwilliams

Here is what the U.S. Copyright Office actually says.

https://www.copyright.gov/ai/

They are releasing it in parts and have only published Part 1 as of today. Part 1 was published in July 2024.

https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-1-Digital-Replicas-Report.pd...

I will have time to read the 72 page document later today. Thanks for making me look this up Namu!

Jane

[D Fosse]

I think it's a safe assumption that "significant human input" goes well beyond typing in a prompt...

[Lumigraphics]

And more importantly, the OP doesn't have the money to chase a copyright infringer in court. Most of us artists are shut out of the court system because of this, even when we have a legit complaint about our own actual artwork being stolen.

[Stephen Marsh]

I am presuming these downloads are from photographs of people in cyberspace, right?

By @Janeames

Adobe generative AI uses training data from Adobe Stock submissions which Adobe has rights to, they don't "scrape the interwebs."