On Thu, Jul 17, 2025 at 07:24:22PM -0400, Simo Sorce wrote: > On Thu, 2025-07-17 at 15:26 -0700, Adam Williamson wrote: > > On Thu, 2025-07-17 at 15:18 -0400, Simo Sorce wrote: > > > In my opinion the situation is simple, as already several courts > > > hinted, the output of an AI cannot be copyrighted, and that makes sense > > > given Copyright hinges on protecting human creativity and AIs clearly > > > are not human. So Fedora could make a decision that the default license > > > for AI generated code is just "Public Domain". > > > > The most 'dangerous' case is the potential one where the AI system's > > output is a close copy of some pre-existing human-authored chunk of > > code, to the extent that the human author's copyright would apply to > > it. In that situation we may be violating their copyright by including > > it, depending on the license. > > Can we talk about what the risk of that is? > Are we talking 30%, 3%, 0.3% 0.003% .. ? > > In the end I would make it a risk calculation, and add a note that if > that should ever happen, and an author comes around and demand changes, > Fedora will expeditiously remove the code or add attribution as the > author prefers. > I assume the chance of that happening to be low of course.
We can hope things are so clearcut & simple. The pain comes when there is a difference of opinion about the merits of the complaint, and/or when the original code in question has become the foundation behind a much larger piece of evolved work so not practically untanglable with alot of collatoral damage. Specifically in the case of Fedora RPM spec packaging work, I'd suggest the risk is likely low. If AI was going to spit out copied work we'd probably find it was copied from another Fedora RPM spec, or copied from examples in Fedora packaging guidelines, and thus not something to particularly worry about. More broadly in terms of coding I'd say the problem mostly falls on upstream projects, as the original coding work in context of RPM package maint is usually fairly low, and mostly around backporting existing upstream work rather than authoring net new code. Other Fedora activities make be a different matter though. Consider documentation authoring, or translation work, where the risk would be cloning paragraphs of text from other sources in the training material. With regards, Daniel -- |: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :| -- _______________________________________________ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/ List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.org Do not reply to spam, report it: https://pagure.io/fedora-infrastructure/new_issue
