Re: GCC AI Policy Announcement

Am 01.08.26 um 21:00 schrieb Sam James:
>> What happens if a patch contains a non-trival amount of code that
>> cannot be copyrighted because it was assisted by AI?
> The open questions around this are why the GNU project issued
> provisional guidance to reject significant contributions developed with
> an LLM.
> 
> If it is copyrightable, then who owns the copyright? If it's based on
> the training data, then the contributor may not have the rights to
> share such a patch, and it cannot be legally redistributed under the GPL
> (or any licence).

That is one of the clearer points right now - LLM-generated code cannot
be copyrighted.

If somebody significantly modifies LLM-generated code, then that person
gains copyright (which would also be fine).

And if anybody ever finds that training data fed to an LLM lead to
copyright on the generated code... this would be such a huge upset
of the software and other industries that anything about GNU software
and gcc would pale by comparison, by so many orders of magnitude
that it would be ridiculous.

It seems that the risk assessment leading to the current policy is
at variance with what everybody else's risk assessment, and frankly,
I do not think that this is justified.

And this policy, if it is held up over a long period of time,
has the potential of significantly holding back gcc's development,
both in absolute terms and in relation to its competitors, for
example LLVM.

I have an e-mail in my inbox with some details of a very worthwhile
project, where gfortran lags behind other compilers by large factors
in speed. There is a LLM-generated prototype which would run
afoul of the current rules, so it seems that avenue is currently
closed. Oh well, then gfortran will continue to lag behind other
compilers.

To put it bluntly: The current policy is based on a risk assessment
that nobody shares, and risks losing significant development speed
over the competition.

Best regards

	Thomas