The way I see it the legal issue of this is not sorted out yet, with several ongoing lawsuits and the agencies like Adobe and Dreamstime that allow AI content stimply trust that the outcome will be in their favor.

And if it turns out to be otherwise they will still not really have lost anything. They might remove AI content again, but I doubt they'd compensate any customers or contributors for any damage, just like it is done with millions of stolen images. Agencies sell stolen content, someone finds their artwork in a port full of stolen images and all the agencies do is remove that single one image and keep the port open, knowing very well that the chance just one stolen image found its way into that port "by accident" is extremely unlikely and all the images are stolen.
But they make money from it, so what do they care? Has anyone ever heard of them contacting a customer telling him "Sorry, you have to stop using that image and we will refund you, because we didn't have the right to sell this image" or have they ever contacted a contributor telling him "hey, we sold this image stolen by someone from you, we have removed it, closed the port and here is the money  the thief made from that sale that should be rightfully yours".  Nope. Not happening. They keep the money even from illegally sold content, so even in the unlikely case that it turns out microstock agencies had no right to sell AI content, they will keep the millions that are rolling it with sales from it now.

Quote from: Her Ugliness on May 06, 2023, 05:31

But they make money from it, so what do they care? Has anyone ever heard of them contacting a customer telling him "Sorry, you have to stop using that image and we will refund you, because we didn't have the right to sell this image" or have they ever contacted a contributor telling him "hey, we sold this image stolen by someone from you, we have removed it, closed the port and here is the money  the thief made from that sale that should be rightfully yours".

But if you were the buyer then you probably wouldn�t want to continue using some stolen imagery you bought. You rather have a refund because, as a buyer of stolen copyright content, you could still face penalties for a rights violation if you continue using the stolen imagery.

Sure, but what does it matter in regards to this discussion? Do you think the customer is ever told by any of the agencies that the image he bought was stolen?

My  point was: Agencies do not care about legal aspects, as long as they make a profit and can get away with what they are doing and I think that's the stance of the agencies that allow AI images while some others don't.

Quote from: NewStocker on May 06, 2023, 05:19

AI Generated Imagery, Is It OK To Sell It?

You can't sell them. This is copyright infringement. You can not take someone else's work and redo it without the permission of the author. And even more so you can not sell this work later as your own.

Tell us you don�t understand AI imagery without telling us.

Sure, but what does it matter in regards to this discussion? Do you think the customer is ever told by any of the agencies that the image he bought was stolen?

My  point was: Agencies do not care about legal aspects, as long as they make a profit and can get away with what they are doing and I think that's the stance of the agencies that allow AI images while some others don't.

No, they do care about legal liability and less about keeping a bit of money that they earned on an illegal stock content sale. So, buyers should be given a refund and told the content they bought does not have legal copyright and that they should stop using it for that reason. But the stock sites usually won't tell them as you said, which is really bad business practice. But it is for different reasons. 

Stock sites give refunds all the time. The real issue in a situation like this is that the stock sites would open themselves up to lots of questions and legal liability if they told a buyer that the content they licensed to the buyer was wrongfully licensed. Questions such as how could the stock site unknowingly be selling content that they don't have correct copyright license on. And then, if the buyer already used the content for a large commercial campaign, the buyer will be worried about their own copyright liability and could even take legal action against the stock site for their blunder. Plus, all around, the optics aren't good with such a clumsy admission like this to a buyer from a stock site.

Anyway, this is all really off-topic and doesn't address the OP's points about why some agencies do license AI generated content and other's don't. Again, I don't think it is about something as simple as trying to make extra profit doing something wrong and getting away with it. The sites that don't license AI content are probably more concerned about potential legal liability down the line because the copyright laws on AI content are still in flux and the stock sites that do license it probably don't think there will be any liability with it later. Different views on it from different stock sites and I think that was part of what the OP was trying to express.

Quote from: stoker2014 on May 06, 2023, 10:26

Quote from: NewStocker on May 06, 2023, 05:19

AI Generated Imagery, Is It OK To Sell It?

You can't sell them. This is copyright infringement. You can not take someone else's work and redo it without the permission of the author. And even more so you can not sell this work later as your own.

Tell us you don�t understand AI imagery without telling us.

Why should I say something? Better than all my words - lawsuits that have already sued websites selling AI-generated content.
I don�t know what kind of responsibility specifically stock agencies will bear, but the �authors� who sell such things may well be made extreme, banned and held accountable.

I am trying to understand the theoretical reasons why some agencies accept AI generated imagery and others don�t. Meanwhile, Adobe is one of the few (and the largest platform) that does accept this type of content for sale. And then Getty (iStock), Shutterstock (Pond5), Alamy, and some others don�t accept it. Or if they do accept it then it has to be imagery created with their own native AI generators that are controlled by that platform.

The basis for not accepting and licensing the content for sale for most of them is that AI image generating models/platforms scrape millions of copyrighted images from the internet in order to source ideas and, thus the content being used to generate the AI imagery is not free from rights.

So is the reason that sites like Adobe, Dreamstime, 123RF and others do accept AI generated content is that they see the AI platforms as only gaining �creative inspiration� and taking �creative license� from all the copyright content that is being scraped by the AI image generators and thus making it a non-violation of usage rights?

I am not sure what the right answer is here, but I guess theoretically these AI models using copyright content to generate their own images would be no different than a photographer looking at an existing copyrighted photo of an apple and then photographing their own apple in the same spirit with similar lighting and composition to the copyrighted image they saw.

An interesting debate.

In theory, an AI which operates in the way you describe is possible. However, to have a full understanding of what is going on under the hood of the popular engines like Dall-e and Midjourney, you'd have to analyze the code they're running. They could as well be taking bits and pieces from multiple images, and mangling them together. This would make the output so obfuscated, that you'd never be able to tell what source images were used.

On top of that, anyone could create a collage of other peoples images, then upload it on a stock agency claiming that it was created by AI.

and the results are new images automatically covered by copyright under US law

Uhm. So far the US copyright law does not agree with you.

...
no need to know their algorithms - as has been shown MANY times here, no images are used in creation of new images...

No they haven't.

Quite the opposite. In fact a link to a white paper was posted here showing that scraped images can be extracted with the right prompt. You are choosing to ignore it.

Not sure what you mean. Of course images are used. How is the AI trained if images aren't "used" to train it.

Quote from: cascoly on May 11, 2023, 09:52

...
no need to know their algorithms - as has been shown MANY times here, no images are used in creation of new images...

No they haven't.

Quite the opposite. In fact a link to a white paper was posted here showing that scraped images can be extracted with the right prompt. You are choosing to ignore it.

Not sure what you mean. Of course images are used. How is the AI trained if images aren't "used" to train it.

you're confusing training & generation - once the images are used in training, the bots do not reference any parts of the original images when generating new art.  references to how th is works have been posted many times if you really want to know how this works

The programmers have appropriated copyright work to make their product, and as proved in the paper I referenced that work is still accessible using the right prompts (though this is not all that relevant as the issue is the appropriation of the work for commercial purposes without the copyright owner�s consent).

Although legally there may be loose details that programmers have infringed upon, the point of AI image creation, as a concept, is exactly the same one that we have all learned to read, write, draw or photograph. I have read dozens of books by cartier bresson, Avedon, Erwit, Leibovitz... and seeing their photos is how I have learned to photograph, and I have created my own photographs with theirs in my mind.
That's the same thing they do, and it can't be avoided, because the photos are published, open to anyone's eye.

Years ago you had to pay to enter the museums, now it's not even necessary.

Quote from: Justanotherphotographer on May 13, 2023, 09:09

The programmers have appropriated copyright work to make their product, and as proved in the paper I referenced that work is still accessible using the right prompts (though this is not all that relevant as the issue is the appropriation of the work for commercial purposes without the copyright owner�s consent).

Although legally there may be loose details that programmers have infringed upon, the point of AI image creation, as a concept, is exactly the same one that we have all learned to read, write, draw or photograph. I have read dozens of books by cartier bresson, Avedon, Erwit, Leibovitz... and seeing their photos is how I have learned to photograph, and I have created my own photographs with theirs in my mind.
That's the same thing they do, and it can't be avoided, because the photos are published, open to anyone's eye.

Years ago you had to pay to enter the museums, now it's not even necessary.

tell that to the musicians who got sued after listening to others work and using it it as "inspiration". 

fascinating how many stock photographer seem to have an international law degree.

And personally I have not seen any cases of plagiarism by AIs yet.

Now you have.

https://gizmodo.com/ai-art-generators-ai-copyright-stable-diffusion-1850060656

Quote from: Yakystockero on May 15, 2023, 03:07

And personally I have not seen any cases of plagiarism by AIs yet.

...
Now you have.

https://gizmodo.com/ai-art-generators-ai-copyright-stable-diffusion-1850060656

what a surprise!  you take a very specific prompt with a tiny training set you esigned & you generate a similar image (which would not be possible with MJ or DALLE since they dont allow such names to be used). 

 the article in the link is highly misleading - when you read the actual article, you discover they did not generate this image from MJ et al - they were on a very specific, biased snipe hunt to prove their hypothesis (not unreasonable since they were trying to show feasibility) -- they used the original images, did their own training, and then generated images with their own generator.-.  their methodology was specifically looking to generate 'memorized' images and is highly unlikely to occur in the real world - while they show it's theoretically possible for this to occur, they haven't shown it does with existing public generators

we still havent seen such results with publicly available

btw, as previously reported, when i tested a very specific prompt for which i had 80% of the images on SS, the result was competely different from the training images.

This has clearly nothing to do with AI images creations, and
It simply seems a low quality jpg, or overcompressed image.

Did you even read the article? How does this have "clearly nothing to do with AI image creations", when the image on the right WAS created with an image AI creator, Stable Diffusion, to be exact?  It's not a low quality compressed image, it is the image Stable Diffusion AI created.

And Stable Diffusion has not been given "one image of a women", but all of the internet. It's currently one of the leading AI image creators out there.
Yes, in this case the people achieving this result gave it a prompt that equaled the exact dataset the AI was trained with, this was not a chance result. No one claimed they entered "blond woman" into the prompt field and by chance got a result that looked exactly like an existing photo. But the point of this is to show the problem that AI HAS the capacity to completely recreate an image that belongs to someone else.

 
Here's the actual paper:

https://arxiv.org/pdf/2301.13188.pdf

When reading bear in mind all the major AI generators now use diffusion models rather than GANs. I described the method before and was told "that's not how AI works" by someone who then proceeded to describe GANs.

The dataset they used to extract images is Stable Diffusion "the largest and most popular open-source diffusion model. This model is an 890 million parameter text-conditioned diffusion model trained on 160 million images."

Here's an extract:
"Examples of the images that we extract from Stable Diffusion v1.4 using random sampling and our membership inference procedure. The top row shows the original images and the bottom row shows our extracted images.
"

Here is what they actually use the smaller data set to investigate (in the second part of the paper):

"The above experiments are visually striking and clearly indicate that memorization is pervasive in large diffusion models�and that data extraction is feasible. But these experiments do not explain why and how these models memorize training data."

So they didn't use a hand picked dataset to prove they could extract the images at all. They used the largest set available. Only subsequently did they use a smaller set to show how they were stored.

 
Here's the actual paper:

https://arxiv.org/pdf/2301.13188.pdf

When reading bear in mind all the major AI generators now use diffusion models rather than GANs. I described the method before and was told "that's not how AI works" by someone who then proceeded to describe GANs.

The dataset they used to extract images is Stable Diffusion "the largest and most popular open-source diffusion model. This model is an 890 million parameter text-conditioned diffusion model trained on 160 million images."

Here's an extract:
"Examples of the images that we extract from Stable Diffusion v1.4 using random sampling and our membership inference procedure. The top row shows the original images and the bottom row shows our extracted images.
"

Here is what they actually use the smaller data set to investigate (in the second part of the paper):

"The above experiments are visually striking and clearly indicate that memorization is pervasive in large diffusion models�and that data extraction is feasible. But these experiments do not explain why and how these models memorize training data."

So they didn't use a hand picked dataset to prove they could extract the images at all. They used the largest set available. Only subsequently did they use a smaller set to show how they were stored.

I read the paper: while aiming specifically at Stable Diffusion (with images based on the LAION dataset) using a specific algorithm  they managed to "extract" 50 images out of 175 million, and all those 50 images were duplicated at least 100 times in the dataset. In order to retrieve the images they had to use as a prompt a string siphoned from the LAION dataset itself. Speaking about doctored stats...

One would conclude that, unless you're specifically hunting for a scandal, the probability of getting a "tainted" image from Stable Diffusion is 1 over 3,500,000.
An average human life lasts (with a bit of luck) about 29,000 days, hence - statistically speaking - you'd have to create with Stable Diffusion 1206 images a day (starting on the day of your birth) before getting one.  Roll up your sleeves... 

The point was to demonstrate that the images are stored in some form which was demonstrated by retrieving them with the right prompt. The relevant point is that the images are, in fact, compressed in a database that is pulled from. I am on mobile so can't  get to the paper right now. Doesn't it conclude that something like 0.03% of the time they were getting recognisable results, so almost 1 in 300? (tough this is really irrelevant, the point is they conclusively demonstrated that original images are stored, how good the app is at covering it up is less relevant).

ETA. Just reread your post. Lol made up stats indeed. You took the total database size and divided it by the 50 images to calculate the probability? Come on, you're a scientist.

Oh, about the stats please teach me! And about the storage, please consider recommending Stable Diffusion not to keep their disks in a mouldy cellar and use the cloud instead