It seems interesting, and we have needed new thinking on licenses for a long while, but a few details mildly concern me.
First, while not universal, brevity improves readability, but often leads to loopholes. Mature licenses grow so long because lawyers have needed to repeatedly game out threats and how to resolve them before arguments break out.
Then, I don’t quite know how to phrase such a thing, but it’d be nice to bar Contributor License Agreements and other copyright assignments. Prior to LLMs laundering Free Software and helping novices flood everybody with low-quality patches, the biggest threat to Free Software was (and probably will be again when the AI bubble pops) was projects exercising the copyright holder’s authority to re-release without a license, taking the community’s work with them behind the paywall.
The LLM condition also seems concerning. I don’t like how the big companies have behaved either, but the problem is corporate exploitation, not the technology. If a teenager wanted to train their own model with a selection of projects, we don’t want to tell them to stop, and if a company wants to repeatedly scrape projects for a PowerPoint deck, that’s just as bad. Plus, training a neural network probably falls well inside the bounds of Fair Use, though publishing the output probably wouldn’t.
Otherwise, you might want to reach out to the folks working on Copyleft Next, which has a similar interest in building on the GPL, since Kuhn and Fontana have been at this for a long time.
Need to come back to this.
Looking forward to comment.
My understanding: this is a whole new class of a license that helps “vibe coding” devs to retains rights/ownership of their code/projects? I mean, crudely put more or less, but anything that’s AI assisted in creation should be retained by that user, right?
Then this helps with people building onto projects that’s aren’t under the same licensing structure?
If I’m not understanding, please assist lol.
License’s focus is not vibe coding devs but in general FOSS devs but current GPL licenses doesn’t protect them from A.I. scrapping their work. This one I wish will protect their works against non FOSS LLM/A.I. models leeching their work.
Ah ok, even better!
current GPL licenses doesn’t protect them from A.I. scrapping their work
The legal status of AI scraping is not dependent on a specific license.
It’s dependent on whether copyright law requires permission from a copyright holder to train AI on their work. This is, as far as I know, not (mostly) a legally settled question yet.
All a license can do is permit things that would otherwise not be permitted. If copyright law doesn’t require that kind of permission, then it doesn’t matter what the author wrote in the license, they won’t be able to successfully sue for copyright infringement.
Meanwhile, if copyright law does require permission from copyright holders for training AI on their works, then the GPL already does what you want to achieve, because then anything generated from such an AI is already a derivative work.
Sounds interesting, although I feel like point 2 is a bit too restrictive.
Right now, it seems to require public availability of the program’s source code, even if the program is only used internally. This, in my opinion, is too restrictive and would make organizations not want to touch software under this license, even if they are friendly to the FOSS community and want to make the end result open source. After all, during internal development, they might be experimenting with lots of different programs/libraries to find the one that best suits their needs, modifying them if necessary, then tossing out whichever one isn’t required. Having to then also put out a public repository during that step feels like too much (and could also hurt secrecy – competitors could see "oh, this company is now publishing a source code for
libfoo, they must be working on somefoo-related project).Perhaps a better rule would be something like “if you deploy the program, you must make the source code available in such a way that anyone who has the technical capability and legal right to use the program can also receive its source code”. This would, of course, mean that while the program is being developed and tested internally, the source code likewise only needs to be shared internally, but when it becomes available publicly, the source code has to likewise be available to any potential user (meaning that someone doesn’t need to, say, buy or own a device running some firmware to get its source code, like with GPL).
Being too restrictive is international. Also I think if organisation using internally it’s difficult to enforce this specific term but it will keep them away of they are not intentional about releasing source code.
