Re: Dealing with LLMs in IETF discussions draft

High level: +1 to most what Rob said.

My thoughts:

a) Rob: "oh look, wasn't it sweet when the humans still thought that they were in control."

   Exactly. Hence (IMHO) the best mailing list name would be [email protected] to
   discuss (use of) AI within IETF itself. Happy to make the ask if some AD/IAB member is
   willing to sponsor.

a) What i am primarily missing is good, community created guidance how to best use
   AI to improve IETF work/process.

   For the time being, i have created:

   https://wiki.ietf.org/group/skynet

   But i haven't found the time to populate with good suggestions floating around
   already on the mailing lists.

   For example i thought to remember someone posted
   pointers to a nice AI method that was trained on all rfc/drafts and mailing list
   discussions - but i can't locate the email anymore *sigh*.
  
c) I don't think we want to rush getting Brians draft published as RFC given the
   ongoing learning about the problems in the community. On the other hand, he wrote
   the text quite well very independent of AI. So i have no hard opinion yet...

d) Side note:

   I would have preferred a wiki page on authors.wiki.com, but that is not community
   maintained, but LLC maintained. Hence it's too cumbersome to quickly add links to
   useful stuff by commuity members IMHO.

Cheers
    Toerless

On Mon, Aug 03, 2026 at 09:28:41AM +0000, Rob Wilton (rwilton) wrote:
> Hi Brian, & Stephen
> 
> Yes, I think that your summary is exactly right.
> 
> I'm not sure what I really think about Stephen & Chong's draft.  I think that I have a couple of concerns:
> 
>   1.
> I'm not convinced that the outcome of the discussion will really be any different from the summary that you have just posted.  I.e., yes, use AI/LLMs, but you are 100% responsible for what you post.
>   2.
> LLM tooling still appears to be advancing rapidly.  It is plausible that in 6-12 months the landscape will have shifted again and whatever gets published as an RFC on this topic will effectively be obsolete by the time it gets an RFC number assigned.  But it could still be a quaint footnote to history, e.g., oh look, wasn't it sweet when the humans still thought that they were in control. 😉
> 
> So, no objections to discussing this somewhere, I may even lurk on a new mailing list to see where the discussion goes.  But I'm somewhat sceptical of whether this is useful work to publish as an RFC.  I would prefer to see it on some IETF webpages, where it could be updated more quickly/easily as the technology evolves.
> 
> But we could bike shed on a mailing list name.  What about "[email protected]", short for "Discussion Is Synthetic; People Left, AI Chatbots Engineer Documents" (*)
> 
> Kind regards,
> Rob
> 
> (*) An LLM may have been involved in suggesting this name 😉
> 
> 
> From: Brian E Carpenter <[email protected]>
> Date: Saturday, 1 August 2026 at 02:50
> To: [email protected] <[email protected]>
> Subject: Re: Dealing with LLMs in IETF discussions draft
> 
> On 01-Aug-26 06:09, Laurence Lundblade wrote:
> > My top-of-mind reaction: people are responsible for what they post.
> >
> > If someone is using AI well and their posts are good, great! If not, we’ll learn that and pay less attention to them.
> >
> > An important guideline: you are responsible for what you post.
> 
> This. You own your spelling mistakes, your technical mistakes, and any LLM hallucinations
> that you transmit.  And it is no excuse to say that the LLM messed up your homework.
> 
>      Brian
> 
> >
> > If you blast out lots of AI-originated text that you haven’t reviewed thoroughly and maybe don’t even understand, your credibility in the IETF is going to decline. At some level, not understanding what you post even becomes a basis for moderators prohibiting posts from you.
> >
> > LL
> >
> >
> >> On Jul 30, 2026, at 6:20 PM, Stephen Farrell <[email protected]> wrote:
> >>
> >>
> >> Hi all,
> >>
> >> Myself and Chong Feng have posted an I-D [1] that (from two
> >> pretty different perspectives) aims to progress discussion
> >> of how we might better handle LLM text in IETF discussions.
> >>
> >> I think this'd be a fine topic for broad discussion. I do
> >> not know what would be the correct mailing list for such a
> >> discussion, but also think it doesn't warrant waiting for
> >> another round of dispatchery before being discussed. (Nor
> >> would any dispatch outcome above 'new mailing list' suit.)
> >>
> >> So, I'd be v. interested in comments on this and/or on it
> >> being moved to some other mailing list for discussion. If
> >> no existing list suits, I'd suggest [email protected]
> >> (which could be contracted to [email protected]:-) as a name
> >> for a new list, but whatever works is fine.
> >>
> >> Meanwhile, we'd be v. interested in any comments here or
> >> on the github repo [2].
> >>
> >> And of course, this is a v. rough thing, not intended to
> >> be prescriptive or authoritative, so have at it:-)
> >>
> >> Cheers,
> >> S.
> >>
> >> [1] https://datatracker.ietf.org/doc/draft-fengfar-led/
> >> [2] https://github.com/sftcd/led/
> >>
> >

-- 
---
[email protected]