Comment by graemep
4 days ago
Its not the real reason. Copyright law varies with jurisdiction, is not simply if its AI written its not covered" anywhere AFAIK, and the and the reasons given in the blog post do have anything to do with copyright.
There is a certain amount of fuzzy thinking, but I think what is significant about this is that Codeberg is not a neutral place for FOSS software. It is a place for FOSS that Codeberg considers good - so no LLM code, no crypto related code, and no doubt more to come. I think it feels odd to people because its a different set of values to FOSS licences which are designed to be apolitical and neutral - a software license that stated that the software could not be used for crypto or in LLM projects would not be a FOSS license.
The most coherent argument I have heard is that LLM output might not be copyrightable. This matches what most courts have said, and is the assumption the entire industry is working under. There are probably eventually going to be cases where the LLM copies something verbatim, and then the courts will have to come up with some refinement to their tests. The legal exposure for this for codeberg is basically zero in the US, thanks to the DMCA.
So, their "copyright status is unclear" stance seems to be saying "it is unclear if the contributor can copyright the code, or if it cannot be copyrighted", but that argument does not make any sense:
"Not copyrightable" means "no one has a right to enforce a copyright on this".
If that is their objection, they need to run scrubs for public domain works + delete those too. God help them if they want to host old UNIX software. Unless someone has renewed the copyrights on stuff written before 1979, the copyright has already lapse, making it "not copyrightable".
If, alternatively, they're worried about inadvertent plagiarism, then they already have to have mechanisms to deal with intentional plagiarism, so no action is required.
If they're trying to impose their values on users by micromanaging acceptable use of third parties' computers, well, I agree: That's explicitly in violation of the definition of Free and Open Source software, so they're basically just ignoring their non-profit mission statement.
That is a good point. You could not host public domain code there, and there are quite important FOSS projects that are public domain - as well as the the old code you mentioned there is current stuff such as SQLite.
> The legal exposure for this for codeberg is basically zero in the US, thanks to the DMCA.
They are in Germany and they distribute software globally (i.e. you can clone or download code from anywhere) so zero liability in the US is not enough.
I think you are being rather harsh on them, but I think your point is more evidence that the real reasons are the ones in the blog, and they should probably remove copyright as a justification in the ToS.
I don't really understand international copyright laws at this point (too many jurisdictions), but, given the number of global sites that can only exist because of the DMCA (like YouTube), I assume there are comparable carve outs allowing platforms to host user uploaded content without being sued out of existence, as long as they pull copyright violations down reasonably quickly.