RFC: The Binutils and LLM generated patches

[Paul Koning]

That policy proposal makes sense to me in general.  In particular, the basic argument why not to accept LLM content is correct in my view.

However, the details seem contradictory.  First it says that LLM content is not acceptable unless "de minimis".  Then it says that LLM content may be included provided the submission is tagged as such.  Which is it?  I think the opening paragraph says it has to be the first (no LLM) rather than the second (LLM if tagged).

paul

> On Oct 2, 2025, at 9:44 AM, Nick Clifton <nickc@redhat.com> wrote:
> 
> Hi Guys,
> 
>  I think that it is time that the GNU Binutils project had a position
>  on whether or not it will accept LLM generated content.  I am
>  proposing the attached text as a description of the project's standing
>  on this issue, and I am interested to hear your comments.
> 
>  My plan is to add this text - or rather whatever final version we
>  reach - to the binutils/README file and the Contributors page on the
>  binutils wiki.  I am also wondering if we should have a file similar
>  to the binutils/SECURITY.txt file.  But if so, what name should we use?
> 
>  The proposed text also invents two new tags for LLM generated
>  contributions:  Generated-By:  and  LLM-Query:  I just pulled these
>  out of thin air, so feel to suggest better ones, or tell me that they
>  are pointless.
> 
> Cheers
>  Nick
> 
> 
>      The GNU Binutils project is currently
>      *NOT* accepting LLM generated patches.
> 
> This is because the copyright status of LLM generated code is
> currently unclear.  The policy applies to all parts of the GNU
> Binutils including, but not limited to, code, documentation and
> the testsuites. 
> 
> There are however some exceptions to the policy:
> 
>  * Using LLMs to assist in writing code is fine providing that the
>    LLM does not actually generate code.  So for example using an
>    LLM for translations is OK.  So is using an LLM to provide text
>    to speech services or to search for information, and so on.
> 
>  * LLM generated code that is not "legally significant"[1] is OK.
>    As a rule of thumb, this means that trivial changes, such as
>    spelling corrections, or small code formatting cleanups are fine.
> 
> Using an LLM to inspire or help create a patch might be OK.  It is a
> question of whether LLM generated output eventually makes it into the
> patch.  If it does, then the patch is unacceptable.  (Unless it can
> be considered legally insignificant).
> 
> If a patch submission *does* contain LLM code then this must be
> clearly indicated.  The identification should take the form of a line
> starting with the "Generated-By: " prefix which identifies the LLM
> used and a line starting with "LLM-Query: " which shows the query
> given to the LLM.  For example:
> 
>  Generated-By: GNU-LLM verson 1.0
>  LLM-Query: Create a program to solve all GNU problems
> 
> Also please note that when submitting a patch to the GNU Binutils
> project you must either use the Developer Certificate of Origin[2] or
> else have completed an FSF Copyright Assignment.  Either way however,
> your submission is an implicit statement that you have the right to
> submit the patch, including any necessary copyright management.
> 
> 
> Footnotes:
> 
> This policy is not set in stone.  It may well be reviewed and changed
> in the future.
> 
> The policy uses the term "LLM generated" rather than "A.I. generated"
> as the later could be misunderstood.  See [3] for more details.
> 
> [1]: https://www.gnu.org/prep/maintain/maintain.html#Legally-Significant
> [2]: https://developercertificate.org/
> [3]: https://www.gnu.org/philosophy/words-to-avoid.en.html#ArtificialIntelligence