With all the discussion around detecting when a code repo contains commits authored by an LLM, I think it is important to note commits like the following in Mozilla Firefox from 2 weeks ago:

"Bug 2011195 - When an agent commits, don't add itself as author"

https://github.com/mozilla-firefox/firefox/commit/71cc24b6a400dbd434e4df37087960d94b764791

I don't think it's a good thing that Mozilla seem to be explicitly encouraging unattributed LLM code in Firefox.

Bug 2011195 - When an agent commits, don't add itself as author r=ai4… · mozilla-firefox/firefox@71cc24b

…dev-reviewers,suhaib DONTBUILD Differential Revision: https://phabricator.services.mozilla.com/D279515

GitHub
@sarahjamielewis according to my understanding of current copyright guidance in the United States, doing this means they forfeit their copyright to the entire Firefox codebase.

@sarahjamielewis @linear no, it just means that the slop output is not copyrightable as a work in itself

(it’s still encumbered by the licences (or lack thereof) of all the stolen works they fed it with)

@mirabilos

my understanding is that what you said is the case if it is disclosed, or failing that as in this case, can be determined after the fact:

"Zarya of the Dawn: A February 2023 decision that AI-generated illustrations for a graphic novel were not copyrightable, although the human-authored text of the novel and overall selection and arrangement of the images and text in the novel could be copyrighted."

but that if it is not disclosed, and the AI-generated output cannot be separated upon request from that made by humans, then the entire work is at risk:

"Théâtre D’opéra Spatial: A September 2023 decision that an artwork generated by AI and then modified by the applicant could not be copyrighted, since the applicant failed to identify and disclaim the AI-generated portions of the work as required by the AI
Guidance."

@linear right. One could argue that the codebase ca. 2019 is likely untainted, and use that as measuring stick for what changed.

The other thing is that you can of course mix uncopyrightable PD stuff into a larger work, if your own stuff is copyrightable. That ofc only works if the former is indeed PD and not encumbered with the rights of the stolen works’ creatives.

@mirabilos yeah. in this case i think it primarily raises concerns for future contributions. one wonders if a judge might, if reviewing this, rule that all contributions after this commit are public domain, unless the provenance of the code is very clear, e.g. patchsets taken from other projects which do practice due diligence