On 3/10/26 8:55 AM, Michał Górny wrote:
Our "AI policy" [2] covers only direct contributions to Gentoo. At the
time, we did not consider it appropriate to start restricting what
packages are added to Gentoo. Still, general rules apply and some of
the recent packages are starting to raise concerns there. Hence I'm
looking for your feedback.
A note: chardet's recent use of LLMs to perform licensing shenanigans is
a specific separate case for me, because I think there are potentially
real legal concerns there (and I doubt many folks on this list would
consider it ethical to "LLM wash" an LGPL code base as was done there).
For the general case though, I have two major concerns with this proposal:
- What counts as "AI slop" or packages using AI? A large number of open
source projects and foundations -- including the OpenInfra Foundation
(OpenStack + kata containers + others) and the Apache Foundation -- have
published specific guidance about how to submit AI generated code. I
doubt we could come to an agreement that would cover all the cases you
care about while excluding the cases I might care to fight for.
- If such a policy existed, it would deputize all Gentoo contributors
into being AI-police, so to speak. It's not my job -- and I have ZERO
volunteer interest in -- policing how other people use AI code in their
own projects. That includes having to evaluate if something is "too AI"
for Gentoo packaging.
If such a policy were to be implemented, depending on the exact
verbiage, I would potentially choose to no longer be a Gentoo developer
than put myself in the place of judging upstreams for whatever choices
they made. I think this is mildly obvious given I'm the maintainer of
our claude-code package, and haven't been shy about my usage of them in
other communities as permitted by policy.
IMO, this primarily has to work out to a maintainer/project decision
point: if someone is uncomfortable maintaining a given package for any
reason, they shouldn't maintain it. I would be extremely +1 to the
python project generally refusing to package chardet 7 (or any other
package which has used *any method whatsoever* to screw developers and
operators out of their copyleft rights).
--
JayF
P.S. While I know I'm not the most active gentoo developer, I do have
lots of experience with my other hats on with combatting AI slop. I've
also had positive experiences with these tools. Slop coming out of an
LLM/AI process reflects on *the people using the tools* as much as the
tools themselves. If there's a project producing crappy code, or trying
to screw old contributors out of their rights -- that's a MAJOR PROJECT
PROBLEM regardless of if AI or LLMs were involved. That's mainly why I
don't like when these conversations are LLM/AI-forward -- we end up
blaming 1s and 0s for actions taken by inconsiderate humans.
