Hi Cayetano, On Wed, 04 Mar 2026 at 08:00, Cayetano Santos <[email protected]> wrote:
> LLMs are impossible to detect, as infringement to a license is: how do > you know if someone respects the license terms of another code ? This is > not a reason not to ask for respecting original authors work, as we usually > do. Maybe I overlook something: the aim of the license is to be able to defend your interests against those who infringes it. Licenses are not some decorum that please us or something we politely ask because it’s “good”. Licenses are about copyright and thus connected to authorship. All that is well-documented and defined in concrete terms, and these concrete terms might have concrete actions in Court. “Using LLM” isn’t defined in concrete terms. What does it mean “using” here? What does it mean “LLM“ here? Etc. I’m not sure to get the parallel. Again, maybe I overlook something. > To me, the parallel holds here: we expect others to respect software > licenses, as we should expect them to respect our claim to explicitly > state the LLM origin of any piece of code. This is the basis. This query could be added to a kind of policy, as discussed here [1]. Well, from my point of view, the ‘AI Tool Policy’ provides a concrete policy that might be some basis. I commented it there [2], feel free to add more. :-) Cheers, simon 1: Re: Can a project accept LLMs' code contribution and remain free software? Ludovic Courtès <[email protected]> Fri, 20 Feb 2026 18:11:26 +0100 id:[email protected] https://lists.gnu.org/archive/html/guix-devel/2026-02 https://yhetil.org/guix/[email protected] 2: Re: Can a project accept LLMs' code contribution and remain free software? Simon Tournier <[email protected]> Tue, 03 Mar 2026 22:48:51 +0100 id:[email protected] https://lists.gnu.org/archive/html/guix-devel/2026-03 https://yhetil.org/guix/[email protected]
