Re: Proposed draft of an AI policy

Norm and everyone,

I like the draft policy, with only a couple of small improvements (of
wording) to offer.

But I have another concern: I think we also need to think about mechanisms
for enforcement (of this or any stated policy). It's better to have the
stated policy than not, but many a good policy has been rendered useless by
a lack of fair process to go with it (where "fair" includes "perceived to
be disinterested and even-handed".)

I can think of two mechanisms (so far):

1. An Editorial Board. This is named to sound more ponderous and scary
than it really is. The EB has no role other than to support a rule that no
PR will be considered without sponsorship from an EB member. (Otherwise the
EB is just a list of names of people who happen to communicate regularly.)
Members could be retired for inactivity and new members could be added by
(unanimous) invitation. The charter members could be our usual suspects.

Giving every PR a formal sponsor inside the group means that every PR comes
with a 'hook' enabling us to pull it for any reason. (Which we have anyway.
But this hook could be used earlier, and also has a pre-emptive effect.) EB
members could sponsor their own *and* one another's PRs to keep things
moving. Their votes would not count differently in the WG, although the
sponsor of any PR, if the only sponsor, would have effective veto power
over that PR, by being able to pull their sponsorship. The PR sponsor's
main function would be to help float and address any and all issues and
concerns related to what I call "para-technical" aspects of a PR (several
of which the policy draft mentions).

I'm glad to hear of any reasons we can't (constitutionally?) or shouldn't
do this, or what hidden gotchas there might be. The core of the idea is to
keep the present honor system for the most active community members while
providing a kind of 'immune response' mechanism for contributions from
outside -- which we still need as they are incredibly valuable.

2. Require every PR to have a brief formal statement listing the tools and
processes used in creating it. That is extra overhead for authors, but the
most regular contributors will have boilerplate. This could be combined
with the EB idea.

Too much?

Thanks for reading,
Wendell


On Mon, May 25, 2026 at 7:51 AM Norm Tovey-Walsh <norm@saxonica.com> wrote:

> Per my action, QT4CG-0165-01, here’s an attempt at an AI policy:
>
> Contributions must not include content generated by large language models
> (LLMs) or other probabilistic tools (including tools which are often
> colloquially categorized as "AI").
>
> This policy applies to the specifications, tests, issues, comments, pull
> requests, and any other contributions to QT4CG.
>
> Although we acknowledge the many widely discussed ethical issues with
> LLMs, we will not refer to those issues as a primary justification for this
> policy. Instead, the specific justifications for the policy are practical