Unable to release application due to too permissive licensing

For legal reasons I license all apps I write with aid from LLMs as Creative Commons Zero. Up until now when I tried to upload a release I thought this wouldn’t be an issue since the Nextcloud app guidelines say:

must contain one of the following licenses (for apps targeting v31 or higher, see > the SPDX License List for details):

   AGPL-3.0-only
   AGPL-3.0-or-later
   Apache-2.0
   GPL-3.0-only
   GPL-3.0-or-later
   MIT
   MPL-2.0


 and:

[the license tag] can occur multiple times with different licenses

Thus I created this in my info.xml:

<licence>CC0-1.0</licence>
<licence>MIT</licence>

But it doesn’t work.

info.xml did not validate: Element ‘licence’: [facet ‘enumeration’] The value ‘CC0-1.0’ is not an element of the set {‘AGPL-3.0-only’, ‘AGPL-3.0-or-later’, ‘Apache-2.0’, ‘GPL-3.0-only’, ‘GPL-3.0-or-later’, ‘MIT’, ‘MPL-2.0’, ‘agpl’, ‘mit’, ‘mpl’, ‘apache’, ‘gpl3’}., line 22

I’m thus unable to release my finished Nextcloud app since the legal uncertainty around LLM-assisted coding points to the likelyhood of me not owning any copyright to sublicense to start with. Thus CC0 must be one of the licenses for the application.

Will I need to stand up a separate Nextcloud app store or might this perhaps be fixed? I do consider the “one of the following” line to indicate that this might simply be a mistake in the validation code.

Since the apps become part of Nextcloud core I think it is intended behavior that the license needs to match Nextcloud core. However this seems to contradict the documentation. could you open this issue in the app store repository so the maintainers can figure out if this is a bug or intended behavior? Issues · nextcloud/appstore · GitHub

Thanks for the answer and suggestion, I’ll do so. But now I’m very surprised by why my app (ExApp, dockerized, hosted somewhere else) would become part of Nextcloud core? :slight_smile: I thought I had read through all documentation but it seems I’ve missed something big.

Well, I personally find it not clear but logical. Each <licence> tag must contain one of the names licenses (where CC0 is not part of the set). You might put multiple <license> tags but each must hold the restriction. So, strictly speaking, it behaves as specified.

The point is (I am not affiliated with NC but just reading the docs and trying to make sense of it) that the appstore only wants to publish such work that will not fall on the company’s feet. That is only open source stuff might be there to prevent juristic backlash.

It might be disputable if CC0 would satisfy this requirement. I’d say one could accept it as well but that is not a lawyer’s opinion.

The other point is that for normal (PHP-based) apps, they are highly interconnected with the core. This, the check (and the motivaton of the check) comes from times where only regular apps were available. With the upcoming ExApps, there is a more separated approach in place that might put some decisions made back then in a slightly different light (regarding the linkage with the core). It does not relate with the self-protection of the company and its appstore.

But one more point regarding CC0: Why would you claim that CC0 is the right one for AI content? I’d say this is a very dangerous assumption (in your way of thinking).

  • If the AI system has no saying in the licensing and has no legal copyright (as a non-human), there are no contract prohibiting you to use it and we are in the nice world where you can define the license of the AI products, yo could put any license on it. Including AGPL, MIT, CC0, or proprietary. Then, you could go with a very strong copyleft if you wanted to be on the sure side or not. All right.
  • If you fear that the current situation is sort of critical and the copyright is not yours. However this might be handled from a lawyer’s perspective, I do not care exactly about this. How do you “dare” to reduce the copyright holder’s rights? By using CC0 you say: Anyone can do anything to this. It is not owned by anyone anymore. I let my own rights vanish. You cannot do this with my work (like the diploma thesis of @christianlupus) and tell everyone they can do anything with it. With a real open source license you would at least put some restrictions on the usage.
    If anyone went suing you, CC0 is a worse decision than anything else. As you allow anyone to do everything with it, the is the maximum damage you can do.

Licensing is not about copying the license of your dependencies. You license your complete work under a certain license (or multiple). You thereby redistribute your dependencies and must obey their licenses. So, your licensing scheme might be restricted and sticking with the official licenses ensures that there is a common sense on what is allowed. But in the end, you publish something under a license and you are responsible if that license is not fulfilling the dependencies. You cannot say “I have no clue”. You are publishing. Period.

Sorry for that last paragraph. It might sound a bit harsh, it is not intended as such. In contrast I just want to make the point clear. If you have a good reference on why CC0 might be a good option with AI content, I would be interested in it as it might change my understanding of this licensing stuff significantly.

Christian

All licenses require that you (the license offerer) have copyright to the thing you license. By setting CC0 on things produced by an AI I have made it clear that I don’t proclaim to have any copyright whatsoever to the content.

The legal issues come if I claim that the content is licensed as (for example) GPL. Now others need to follow that license or be in breach of my copyright, and they in turn rely on their additions also being GPLd. If courts now find that AI generated content isn’t copyrightable those who has based their work on believing they have a proper license no longer do so - and their work in turn also becomes public domain.

(I’m not a lawyer but I have extensive corporate experience from the software side regarding IP rights in cooperation with actual Legal)

On the topic of why Nextcloud imposes restrictions at all regarding ExApps I agree with you - there’s no legal reason for doing so, I assume it’s just ideological.

I’d say otherwise.
By publishing under CC0 you are in fact abstaining from your copyrights (at least im American right AFAIK). (In Germany, for example, you cannot do this and CC0 is invalid. But that is a side topic.)
You are not (clearly) proclaiming to have no rights. If that was the case, you would not even be allowed to publish it at all.

All right, I agree with you.

Although, the question would be: Who has the rights on it then or is it without owner. A question for the courts. It makes a small difference on who might have an interest in suing you or your dependees.

This is something, I would not agree with. If I for some reason could get some proprietary data from any company and released that to the public under a permissible license, the company could for sure sue me. No discussion. The data I leaked would for sure not be CC0 from that point on. They would (rightfully) try to stop it spreading.

And by the way: Whose problem is this now?
It is yours. You published a work under a permissible license and others trusted in that license. If that license was illegal for some reason and thus is invalid, they might come at you.

I agree that this would be a complete mess in total. Courts would be bound with this for years, I guess to sort everything out in every relevant country. So, I get the point that you want to be on the safe side.
But, hey, you use AI. As we do not have juristic settlement yet, there is simply a risk involved


You seem to believe that an item that’s not under copyright somehow is illegal to do things with. It’s the opposite. If something isn’t under copyright then it is in the public domain.

By publishing under CC0 I’m letting everyone know that they’re in the clear regardless of what stance the courts take. I’m also in no risk myself of having made claims I don’t have the legal right to.

Quite contrary. But


At least in German law, everything crafted by a human (directly or indirectly) has the undeniable and irrevocable copyright called Urheberrecht. There is no way in German law to have something under public domain. You can grant a license to the public to everything you want but you cannot abstain from that right. CC0 (as a license) tries to mimic public domain.

What I try to say: When we discuss juristic issues, we have to be careful if this is national law and where it applies.

Apart from that, I see that we do not agree on this topic. For me that is fine and I will avoid a discussion battle. It might depend on what you want to achieve (or what you fear most) what a valid way to go is.

Christian

No, not indirectly. The whole point is that AI assisted development might not be considered “crafted by a human” and as such isn’t protected by copyright. Without copyright you (the human) cannot license the work. If you do, other who relied on your claim (not knowing that you had used LLMs, for example) might have the status of their works affected by such a court ruling.

“Only the author’s own intellectual creations constitute works within the meaning of this Act.”