Re: LLM Commit in upstream `master`?!

On Tue, Jul 28, 2026 at 06:56:30AM +0200, Dr. Arne Babenhauserheide wrote:
> Stefan Monnier <[email protected]> writes:
> 
> > The problem is both the force push itself and more importantly the way
> > this was handled, over-dramatizing this very minor issue and rushing to
> 
> I do not consider it a "very minor" issue, but both a huge risk of
> copyright litigation (I expect that there are many lawyers who would go
> after Emacs at the first chance) and of losing trust (not just here:
> there are many outlets who go against GNU when there’s any opening).

Note that after a revert, the offending piece would be buried somewhere
in the git history, not anymore part of the "officially distributed"
Emacs code.

Now I'm sure that there is no legal precedent at all on whether a tidbit
in git history alone would be able to infringe a copyright.

At this point you are multiplying two very low probabilities (that the
LLM infringing copyright by "hallucinating" a commit message alone -- I
believe Yuan in this -- is enough for an infringement, given all the
complexities of proving "derived work" /times/ that the whole repo is
considered copyright-relevant).

> So I’m thankful that this was handled swiftly, though technical
> limitations of git don’t enable a fully clean solution.

I'm clearly with Stefan on this. Seems that there are two opinions :)

Cheers
-- 
t

-----BEGIN PGP SIGNATURE-----

iF0EABECAB0WIQRp53liolZD6iXhAoIFyCz1etHaRgUCamg5XAAKCRAFyCz1etHa
Rm+cAJwMJqn2oHLIgLsi/FyWIxGDhKZTbgCffIjc29hv576iwHdCK/rVQvFypbY=
=NByF
-----END PGP SIGNATURE-----