Re: Ballot option: Call the Free Software community to develop Free AI.

Adrian Bunk <[email protected]> Wed, 29 Jul 2026 13:11:40 +0300
Newsgroups gmane.linux.debian.devel.vote
Message-ID <amnR3KqyDCDbiw1l@localhost>
On Wed, Jul 29, 2026 at 10:21:08AM +0100, Alastair McKinstry wrote:
> 
> On 27/07/2026 08:35, Adrian Bunk wrote:
> > On Mon, Jul 27, 2026 at 02:02:14PM +0900, Charles Plessy wrote:
> > > ...
> > > indeed, a statement that generative AI should not be used against the authors
> > > of its training data better conveys what I tried to express in my draft.
> > The right to fork and compete with the fork against the original authors
> > is pretty central to our values.
> Has there been any legal rulings or official stances (eg of FSF) on AI
> models being derived works if they are trained on GPL code?

AFAIK no.

Sean mentioned that the FSF is currently consulting with lawyers to 
write an LLM contributions policy for GNU.

Ongoing court cases in the US have either been settled or are in earlier 
phases like discovery.

There is an ongoing case at the Court of Justice of the European Union[1]
which includes the question whether training counts as reproduction
under copyright law.

Note that the above is about the second bullet point in the GR proposal 
from Charles.

My problem is with the first point, where I do not see how to interpret 
that other than us affirming that authors of GPLed software should have 
a right to ban using their source code in AI training (even if the result
would be GPL). Such software could not be in main due to not being DFSG-free.

> Thanks
> Alastair

cu
Adrian

[1] https://eur-lex.europa.eu/eli/C/2025/3039/oj/eng