I took the title in a shock until I see it does not include "upstream
projects using LLMs for development" and "AI software".
To me the core of the proposal is trust issue and potential gap of
auditing. From a research perspective I speculate it is too early to
draw a firm conclusion on a quickly moving target (the quality
rationale). For example, are we sure the claims about LLMs'
incapabilities can still hold after 2 years? The situation looks like we
are going to ban a tool because there are trust issues (including legal,
stability, LLM output auditing, etc), while the real issue stems from
people who are potentially going to use the tool in unexpected ways,
instead of the tool itself. -- Just because knife can kill so we ban knife?
The gap of auditing capability for the underlying human is not specific
to Debian, but general to the human, where specialists are able to audit
the generated work, while new people continue to offload their own
effort to LLMs, eventually leading to the situation where there is
nobody on the way of becoming a professional. For example, under the
current trend, the number of people who learns coding the hard way is
decreasing. This is not a countering GR proposal. I understand the GR
proposal, but I hold my doubts at the same time.
On 7/22/26 8:35 AM, Matthias Geiger wrote:
Hi all,
What follows is a GR proposal to ban LLM contributions.
It is time for Debian to make a statement regarding generative AI and
LLM usage.
Since this has been discussed in exhaustive detail on -private and
-project,
the proposal follows below without any further discussion.
BEGIN PROPOSAL
Preamble
===========
This proposal aims to expressly forbid any contributions to Debian
written with
the use or assistance of large language models (LLMs) or other
generative AI tools.
The scope of this GR is (non-exhaustive): - Debian source packages
- Official Debian project software, such as lintian
- Debian web resources - Documentation and translations added by
Debian contributors
- Official communication from Debian
It does not include: - Upstream projects using LLMs for development
- AI-related software
Rationale
===========
[...]
