Actually this (and ONLY this) is what Midjourney ToS says about it:
Subject to the above license, you own all Assets you create with the Services.
IMO there is no takeaway. Wild west situation basically.
This sounds right, but is very complex. Let’s say I draw a picture and I use Photoshop’s Neural Filters or take a photo and use content aware fill - I’m already using AI generative tools. Plus I think Adobe is actually working on their own image generation thing… and there is Stable Diffusion plugin for Adobe already out there…
But let’s say if Epic does manage to somehow specify it… How are they going to enforce that rule? I can recognize most of Midjourney generated images due how that system is biased, but I bet someone that didn’t work with it - couldn’t say it on first glance. But with Stable Diffusion the situation is even worse - for some of the images I generated, even that I know they are AI generated - I wouldn’t recognize them as clearly AI generated.
And those systems will get even better at generation, so maybe in a year it will be even more difficult.
Me too.
Okay, in a legal document, the entire document applies — you cannot just pick bits and pieces.
Rights you give to Midjourney
By using the Services, you grant to Midjourney, its successors, and assigns a perpetual, worldwide, non-exclusive, sublicensable no-charge, royalty-free, irrevocable copyright license to reproduce, prepare Derivative Works of, publicly display, publicly perform, sublicense, and distribute text, and image prompts you input into the Services, or Assets produced by the service at your direction. This license survives termination of this Agreement by any party, for any reason.
In other words, you may create an image using a bunch of words, but if someone creates the exact same image Midjourney has the right to assign that image to someone else.
Your Rights
Subject to the above license, you own all Assets you create with the Services. This does not apply if you fall under the exceptions below.
Yes, you own them BUT the previous paragraph also applies. Yes, that’s how they get around it. So you can own it, but it’s not exclusive ownership.
Please note: Midjourney is an open community which allows others to use and remix your images and prompts whenever they are posted in a public setting. By default, your images are publically viewable and remixable. As described above, you grant Midjourney a license to allow this. If you purchase a private plan, you may bypass some of these public sharing defaults.
Okay, I slightly misremembered what this text said (my bad). Basically what they are saying is you have no EXCLUSIVE rights to the images (see rights you give to Midjourney). Yes, you own them insofar as they also can hand out the same image to thousands of other people because by accepting this you literally give them that right.
That’s you granting them license, not them granting you license.
Not really, it’s just not gone to court. Courts typically like to go the easy route and will either decide 1) no one owns the rights to the images or 2) the image is a derivative of the text entered.
I think they will go with 2 because going with 1 is problematic.
Doesn’t matter what you and I think about it wrt to how laws are being set and enforced about it (unless you are in a position of deciding those laws in your country 
). And btw. in many countries law don’t work like in US. There legislative branches of government decide about laws, not courts alone based on singular cases.
At least with most traditional marketplace content like prop models, you could reasonably expect that a marketplace asset is the exclusive property of the individual/business entity that created it. Sure, stock imagery can be incorporated into an artwork, but most of that action is limited to decal materials where common identifiers are found, like street signs.
If someone is ripping off your game, it’s much easier to make a case with traditional assets that you own a license and the offending party might not.
Wild west indeed.
1 Like
…and this is true for now… Till 3d models generation will become a thing. At that moment the marketplace will be flooded with them too. And that is not that far off. And unless something significant will change in how fast governments are typically setting laws - there will be no clarity for those too.
1 Like
And I will be one of the first to flood the Epic marketplace with AI generated 3D models. Because appearently Epic doesn’t care either, so why should I ?
3 Likes
Why shouldn’t you? It follows that if there’s an easy/low-effort way to make money, people will do it. I can’t blame anyone for that.
I still think the market is better served by keeping extremely low-effort stuff off the marketplace, but that’s not my decision to make. I can only comment on this from my perspective as a frequent buyer. The more the market is filled with this sort of stuff, the less interest I have in sorting through it all to add packs to my unrealsales.io list for potential future purchases.
4 Likes
Let me quote first sentence of Marketplace Guidelines:
Ironic, isn’t it?
3 Likes
I believe that this topic has significantly diverged from the OP concern regarding copyright. I believe I have adequately pointed out that this concern doesn’t seem to be an issue as ownership of the object (aka art) in the case of Midjourney, appears to be the authors and, according to the IP/copyright/patent lawyer, this can be considered a derivative work (for now) because it requires the input of a user to create the artwork (again, they even make the point that it doesn’t HAVE to be good or even valued to be copyrightable).
Next, is the issue of ownership. Since you have to pay-to-play, I see no issue (again) with ownership — a very important aspect for both Epic and people who license such artwork.
Next, there is a confusion about “Is it AI Art?” and that, again, the lawyer explains, no, because AI art that is not copyrightable is art that has no input from a human being. Midjourney, DALLE, et. al. require at least a minimum of user input (again, quality of input is not an issue). This is wholly different from the lawsuit where they said the image “was autonomously created by a computer algorithm running on a machine” (US Copyright Ruling on AI Authorship) — again, we are not talking about the absence of a human author, in fact, they state, explicitly, "The Board accepts as a threshold matter Thaler’s representation that the Work was autonomously created by artificial intelligence without any creative contribution from a human actor” (emphasis mine).
This is extremely important because, as the video above supports, autonomous artworks have no human actor and therefore cannot be copyrighted, however a case could be made that a human entering text that generates art could be considered a derivative work.
It is hardly the wild west (I remember when software copyrights were up in the air!). For Epic, selling things through their stores requires ownership. Whether the person selling created it or not is irrelevant, as is copyright (although, as I indicated, I believe the copyright will be extended to consider these as derivative works).
These images certainly are not autonomous works of computers and therefore the ruling handed down to Stephen Thaler do not apply.
1 Like
I could be misinterpreting things here, but I think it’s too early to say that the copyright issue is adequately resolved. I think it would be more accurate to state that this issue will likely be resolved in the future following the logic of your statement.
Sites like Getty and Shutterstock are reportedly banning submissions containing A.I. generated content over the copyright issue. The issue must be relevant if these sites are taking action over copyright concerns.
The other issue here is that these tools generate imagery by mining content without authorization from their respective creators. Arguably, all artists take inspiration and thus ‘mine’ other artworks to some extent. However, this issue may factor into the copyright problem. If the A.I. generator samples content, then it’s a derivative work of that original artwork, no? If it’s a derivative, should the generator only be allowed to sample from authorized works? Should it be required to give accreditation? Is the original artist owed any compensation?
Is it possible all this impacts the license granted by Epic to these works? I don’t know.
I’ve made several comments on various market items warning others about the issues surrounding ownership of A.I. generated artwork. I’m going to remove these comments as I now realize they contained inaccurate statements.
2 Likes
I have looked at Shutterstock’s submission guidelines (Shutterstock Submission Guidelines) and see nothing regarding AI art. In fact they state specifically:
You must own or control the copyright to all content you submit to Shutterstock. This means that you cannot submit work obtained from other sources (e.g., online image search results or websites), or incorporate such work into your content submissions, unless you have permission to do so. — Shutterstock Submission Guidelines
Their document is quite clear:
- I most own the content
OR
- control the copyright
HOWEVER they also state with this catch-all clause:
We reserve the right to remove any previously accepted content submission from the Shutterstock library, and reserve the right to reinstate any previously accepted content submission which was later removed by a Shutterstock administrator, at any time for any reason.
So, the reality is that you can follow the rules to the letter with Shutterstock and they can remove images because of course they can, no reason required. Why? Because they told you they can.
The downside to this kind of “oh, here are the rules and by the way, we don’t have to follow them” devolves into corporations making arbitrary decisions with no arbitration. I may have a perfectly acceptable photo of a family at dinnertime, and they may strike it because they are interracial and one has a rainbow pin on them.
Do they have to tell me why? No, because they just said they can remove anything for any reason at any time. I don’t do business with Shutterstock for that and other reasons.
The other issue here is that these tools generate imagery by mining content without authorization from their respective creators.
Yes, I agree, as an artist this does give me pause if and only if the artists did not give authorization. Derivative works from works that are public domain, do not require attribution.
As always, when dealing with capitalism, let the buyer beware (I’m aware this appears to be a jaundiced view, but even I have to accept the reality)
Is it possible all this impacts the license granted by Epic to these works? I don’t know.
As someone with a Midjourney membership, I feel confident that I can use artwork created by it without worry up and until it is ruled that the art violates someones ownership. Predicting the future is difficult and I’m not really concerned ATM.
From Epic’s point of view, I believe they have consulted with lawyers who have pointed out that there is nothing to suggest the lack of ownership. Case closed (until ruled or legislated otherwise).
As I have indicated, what is important to business is the trading in licit goods. Epic asked @FaithBasedGames regarding proof — in doing so it was doing it’s due diligence in assuring that he indeed own the images. At the moment, they do — whether Shutterstock or Getty believe they do is irrelevant.
I think where your concern is that can Midjourney assign ownership to a composite image generated by someones word choices?
At the moment, they can.
I could be misinterpreting things here, but I think it’s too early to say that the copyright issue is adequately resolved.
Agreed. However, at the moment, ownership is resolved until is is either ruled or legislated otherwise.
I do plan on using AI art for items that are easily replaced. My reality is I place low value in the images in case a ruling or law changes ownership.
Interesting that they are “reportedly” doing it when in reality that very image can be found on duh duh duuuum Shutterstock AI Generated Image.
Either the writer of the article never bothered to search Shutterstock or they just made it up. I searched for both Midjourney and found 286. Then I searched for ai-generated and found "24,948 ai generated stock photos, vectors, and illustrations are available royalty-free.”
So someone didn’t do their due diligence and wrote an article (Getty may have removed them, or people just aren’t saying they were AI generated).
Shutterstock just banned AI generated images from others…
(the information about banning was sent to content creators by e-mails btw. just search reddit for some people showing those e-mails)
That’s the root of the problem IMO. You can potentially recognize a lot of Midjourney images due to their style… unless you use --testp… but outside of some obvious inaccuracies (like number of fingers, or eyes) you cannot really say with 100% certainty that image was generated or not. And the systems will be improved to remove those inaccuracies.
Interesting, as when I looked at OpenAI license, the following jumped out at me (as always, emphasis is mine):
As between the parties and to the extent permitted by applicable law, you own all Input, and subject to your compliance with these Terms, OpenAI hereby assigns to you all its right, title and interest in and to Output. OpenAI may use Content as necessary to provide and maintain the Services, comply with applicable law, and enforce our policies. You are responsible for Content, including for ensuring that it does not violate any applicable law or these Terms.
Part 1 was the first red flag for me: to the extent permitted by applicable law — so, you are offering this without even understanding or bothering to determine if it is legal? A big nope.
Part 2: You are responsible for Content turns it around and says that they are not responsible for generating copyright infringing material — that’s my problem.
I’m aware that a business must make a profit, but when you start out saying, “Hey, we don’t know if this is legal, and if it isn’t then you need to pony up and fight the legal fight, it’s not our problem” — just seems sketchy.
That’s the root of the problem IMO.
As I pointed out, OpenAI (when I first investigated it) was already illegally using other people’s works (I’ll say allegedly just to be nice). But the agreement was what immediately soured me.
1 Like
OpenAI says this about the data:
DALL·E 2 was trained on pairs of images and their corresponding captions. Pairs were drawn from a combination of publicly available sources and sources that we licensed.
Source: https://github.com/openai/dalle-2-preview/blob/main/system-card.md#model-training-data
One can interpret ‘publicly available’ both ways.
Stable Diffusion was trained on huge amounts of data scrapped from the web ( LAION-5B). Which is using copyrighted materials without consent of copyright holders. The only thing on data for Midjourney I found is this on their discord:
Source: Discord
Which again can be interpreted both ways. But negative interpretation is that they also use copyrighted material without permission (LAION).
I had both Stable Diffusion and Midjourney generating me images with (illegible) watermarks and signatures.
It is 2d images and concept arts for now. I wonder what would happen when people start releasing AI which’s been fed with 2D to 3D art and can spit out 2d concept images + 3d meshes out. Will the work i do still be counted as a job? Or will i be the middle man who writes correct words to get meshes i want and convert them into 3d asset packs? Would people still buy those packs while they can use the AI themselves to get their 3d art themselves?
I always thought machines will be taking the “muscle work” from people; but instead they are doing creative work of the humans… that’s hilarious, and sad. Appearently we will be the ones who will do the muscle work for the machines after all.
3 Likes
You mean like this:
It’s not as good as the original one from Google (or Nvidias one they recently showcased the demo)… but it’s already out…