The fact that AI code is not copyrightable under current rules is super important. You can sit there and smugly attack a project for taking a hard line on this, but if coders using llms end up making open source projects uncopyrightable, then they end up effectively eliminating the enforceability of the open source license.
It isn't about whether AI is a useful tool or not. It is about whether a single bad actor can effectively steal a project from its original authors. The same concern potentially applies commercial works created using AI as well. In that sense, a lot of people are playing dice with the devil.
It isn't about whether AI is a useful tool or not. It is about whether a single bad actor can effectively steal a project from its original authors. The same concern potentially applies commercial works created using AI as well. In that sense, a lot of people are playing dice with the devil.
- replies
- 3
- announces
- 1
- likes
- 6
@sj_zero
Considering that the code produced by AI is by virtue of it's training data, largely derivative from various FLOSS projects, this seems like a chicken-or-egg question. Secondly, if developers are running their models locally rather than using commercial surveillance crapitalist models hosted by the AI slop factories, then no one will be the wiser in any case.
Considering that the code produced by AI is by virtue of it's training data, largely derivative from various FLOSS projects, this seems like a chicken-or-egg question. Secondly, if developers are running their models locally rather than using commercial surveillance crapitalist models hosted by the AI slop factories, then no one will be the wiser in any case.
@sj_zero Super important for the "Free" software community, which uses copyright as "copyleft" to enforce sharing rules.
I've only seen permissive "Open" software licence people worry about the usual suspects taking their code and GPL licensing it, and some wanting to limit the ability of big companies to wield software patents as weapons, thus the Apache licence having a clause about that. But I still see most Open projects using the BSD or MIT licences.
It actually isn't a chicken or egg question, legally. The problem is that the code developed by the AI itself isn't copyrightable. Things produced by AI have been confirmed as not copyrightable (at least in the US), and that's held up in court so far.
https://media.cadc.uscourts.gov/opinions/docs/2025/03/23-5233.pdf
https://media.cadc.uscourts.gov/opinions/docs/2025/03/23-5233.pdf
@sj_zero
Legal or not, that seems bullshit to me. Firstly, it's derived from FLOSS. Secondly, if it comes from a commercial model, then it's a product which was paid for by the consumer, who therefore owns the code. That said, if people knowingly spend their money supporting that kind of bullshit, they'll get what they deserve. However as you said, it could also harm legitimate FLOSS projects if unscrupulous contributors submit patches which are then unwittingly accepted under these legal conditions.
Legal or not, that seems bullshit to me. Firstly, it's derived from FLOSS. Secondly, if it comes from a commercial model, then it's a product which was paid for by the consumer, who therefore owns the code. That said, if people knowingly spend their money supporting that kind of bullshit, they'll get what they deserve. However as you said, it could also harm legitimate FLOSS projects if unscrupulous contributors submit patches which are then unwittingly accepted under these legal conditions.
@sj_zero Sounds like we need a hillbilly from Georgia with a fiddle