Dealing with LLMs in IETF Discussions
draft-fengfar-led-01

Network Working Group                                         S. Farrell
Internet-Draft                                    Trinity College Dublin
Intended status: Informational                                   C. Feng
Expires: 7 February 2027                                   6 August 2026

                 Dealing with LLMs in IETF Discussions
                          draft-fengfar-led-01

Abstract

   The rapid adoption of AI language tools has prompted concern across
   professional and technical communities, including the IETF, about
   authenticity, accountability, and the integrity of human
   contribution.  This document approaches the question from two
   directions: a critical reader's concerns about what AI use means for
   IETF discussion, and a practitioner's account of how AI is currently
   being used in IETF discussions.  We aim to explore some of the issues
   arising, and perhaps make some specific (but tentative)
   recommendations, but the main recommendation is that the IETF should
   develop guidelines for use of AI tooling when engaging in IETF
   discussions.

Discussion Venues

   This note is to be removed before publishing as an RFC.

   Source for this draft and an issue tracker can be found at
   https://github.com/sftcd/led.

Status of This Memo

   This Internet-Draft is submitted in full conformance with the
   provisions of BCP 78 and BCP 79.

   Internet-Drafts are working documents of the Internet Engineering
   Task Force (IETF).  Note that other groups may also distribute
   working documents as Internet-Drafts.  The list of current Internet-
   Drafts is at https://datatracker.ietf.org/drafts/current/.

   Internet-Drafts are draft documents valid for a maximum of six months
   and may be updated, replaced, or obsoleted by other documents at any
   time.  It is inappropriate to use Internet-Drafts as reference
   material or to cite them other than as "work in progress."

   This Internet-Draft will expire on 7 February 2027.

Farrell & Feng           Expires 7 February 2027                [Page 1]
Internet-Draft            LLM Email Discussions              August 2026

Copyright Notice

   Copyright (c) 2026 IETF Trust and the persons identified as the
   document authors.  All rights reserved.

   This document is subject to BCP 78 and the IETF Trust's Legal
   Provisions Relating to IETF Documents (https://trustee.ietf.org/
   license-info) in effect on the date of publication of this document.
   Please review these documents carefully, as they describe your rights
   and restrictions with respect to this document.  Code Components
   extracted from this document must include Revised BSD License text as
   described in Section 4.e of the Trust Legal Provisions and are
   provided without warranty as described in the Revised BSD License.

Table of Contents

   1.  Introduction  . . . . . . . . . . . . . . . . . . . . . . . .   3
   2.  A Reader's Concerns . . . . . . . . . . . . . . . . . . . . .   3
   3.  One Author's Workflow . . . . . . . . . . . . . . . . . . . .   5
     3.1.  Monitoring and Triage . . . . . . . . . . . . . . . . . .   5
     3.2.  Forming a Position  . . . . . . . . . . . . . . . . . . .   5
     3.3.  Drafting in English . . . . . . . . . . . . . . . . . . .   6
     3.4.  Final Review and Send . . . . . . . . . . . . . . . . . .   6
     3.5.  The Role of Odyssey . . . . . . . . . . . . . . . . . . .   6
   4.  Human and AI: Complementary Capabilities  . . . . . . . . . .   6
     4.1.  What AI Does Well . . . . . . . . . . . . . . . . . . . .   7
     4.2.  What AI Does Poorly . . . . . . . . . . . . . . . . . . .   7
     4.3.  The Symmetry  . . . . . . . . . . . . . . . . . . . . . .   8
     4.4.  A Collaborative Paradigm  . . . . . . . . . . . . . . . .   8
       4.4.1.  The Core Principle  . . . . . . . . . . . . . . . . .   8
       4.4.2.  Transparency as a Norm  . . . . . . . . . . . . . . .   9
       4.4.3.  The Deepening Relationship  . . . . . . . . . . . . .   9
       4.4.4.  What Becomes Possible . . . . . . . . . . . . . . . .   9
   5.  Risks . . . . . . . . . . . . . . . . . . . . . . . . . . . .   9
   6.  Initial Discussions . . . . . . . . . . . . . . . . . . . . .  11
     6.1.  Possibly Relevant Policies Elsewhere  . . . . . . . . . .  11
     6.2.  Other points  . . . . . . . . . . . . . . . . . . . . . .  11
   7.  Recommendations . . . . . . . . . . . . . . . . . . . . . . .  13
   8.  Conclusion  . . . . . . . . . . . . . . . . . . . . . . . . .  14
   9.  IANA Considerations . . . . . . . . . . . . . . . . . . . . .  14
   10. Security Considerations . . . . . . . . . . . . . . . . . . .  14
   11. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . .  14
   12. References  . . . . . . . . . . . . . . . . . . . . . . . . .  14
     12.1.  Normative References . . . . . . . . . . . . . . . . . .  14
     12.2.  Informative References . . . . . . . . . . . . . . . . .  15
   Appendix A.  Change Log . . . . . . . . . . . . . . . . . . . . .  15
     A.1.  Draft-00  . . . . . . . . . . . . . . . . . . . . . . . .  15
     A.2.  Draft-01  . . . . . . . . . . . . . . . . . . . . . . . .  15

Farrell & Feng           Expires 7 February 2027                [Page 2]
Internet-Draft            LLM Email Discussions              August 2026

   Authors' Addresses  . . . . . . . . . . . . . . . . . . . . . . .  15

1.  Introduction

   "IETF discussions" here includes emails sent to IETF lists,
   presentations (slides) used at meetings, text input during e.g.
   github issue or PR discussions, and potentially messages sent using
   IM tools.  Text included within Internet-drafts and RFCs is not
   included in scope here, even though some of the same issues will
   arise.  We omit those as Internet-drafts and RFCs are also covered by
   BCP 78 [RFC5378] and BCP 79 [RFC8179] so additional considerations
   apply for such text.

   AI language tools are now widely used in professional writing,
   including by some participants in standards development communities
   such as the IETF.  This has produced at least two kinds of reaction:
   uncritical adoption, where AI output is used where previously a
   person would have written an email, and skepticism, where messages
   bearing the appearance of AI involvement are seen as problematic, on
   the basis that a reader cannot tell whether it is the person sending
   the email or just the AI tool, or some mixture.

   In order to explore these positions, it may be helpful to outline
   them in more detail, with the goal of better understanding what AI
   does well, what it does poorly, and where the boundary between them
   lies.

   This document grew out of a specific exchange on an IETF mailing
   list.  One author described using AI to help express ideas developed
   independently; a reader flagged the output as LLM-generated and
   disengaged.  Neither was wrong.  But the exchange exposed a gap: the
   community lacks shared norms for how AI assistance should be used and
   disclosed in email discussions.  Rather than treat this as a local
   disagreement, the two parties decided to try think through it and to
   document that discussion.

2.  A Reader's Concerns

   This section is written by the first author.  But readers may likely
   have guessed that anyway:-)

   Current AI tooling tends to emit text that can be readily seen to
   have involved that tooling.  The following seem to be some of the
   current "tells" for AI having been used when one considers the stream
   of email messages arriving from a sender:

Farrell & Feng           Expires 7 February 2027                [Page 3]
Internet-Draft            LLM Email Discussions              August 2026

   *  Frequently being overly positive about a message to which this
      message is a reaction, e.g. "You've asked exactly the right
      question..."

   *  Unexpected/over-use of geometric terms, e.g. "The seam is..." or
      "There are 17 dimensions..."

   *  Specific phrasing patterns, e.g. "Fifteen wibbles: two designs."

   However, a perhaps more disturbing pattern is the lack of
   uncertainty.  It seems that people using AI tooling don't ask others
   what they mean, perhaps as AI tools make a statistical choice as to
   the meaning of earlier messages, then react as if that is a given.
   It's hard to see how that cannot lead to radical misunderstandings
   and, given AI tooling imperfections, senders emitting relative
   gibberish.

   Use of AI tooling also seems to correlate with "walls of text" that
   are very difficult to parse, both due to length (or seeming
   completeness), and complex sentence structures.

   Readers of such messages also generally have no insight into the
   tools used by senders, nor the level (if any) of human pre- or post-
   processing of AI inputs and outputs.

   In some cases, such messages may be sent in a time-frame that would
   seem impossible for a purely human-generated message, which also
   decreases confidence in the level of human input involved.

   All of the above means that a reader who considers that a message (or
   stream of messages) is largely the output from AI tools can have no
   confidence that they are really discussing a topic with the person
   who seemingly sent the email.  At that point, the only rational
   action seems to be to ignore such messages as being equivalent to
   spam.

   Note that the above issues are not the same as the sock-puppet
   problem, or a sybil attack.  These issues remain problems even when
   the sender of messages is known to be a real person engaged in IETF
   work.

Farrell & Feng           Expires 7 February 2027                [Page 4]
Internet-Draft            LLM Email Discussions              August 2026

   Despite all the above, readers do know that AI tools are being used
   and have to be dealt with, and that ignoring messages won't scale if
   use of those tools becomes more common, nor if the tools get better
   to the point where messages no longer expose use of such tools.  And
   it has to be acknowledged that the translation capabilities of AI
   tools could be beneficial to the Internet community, in terms of
   opening up participation to many more capable engineers for whom
   communicating in English is a challenge.

   It therefore seems possibly useful to explore these issues in more
   detail, hence this draft.  (This author does not expect this draft to
   eventually become an RFC.)

3.  One Author's Workflow

   This section is written by the second author, who uses AI assistance
   in IETF participation.  While this author's workflow does envisage
   use of AI tooling for Internet-draft and RFC text preparation,
   dealing with that aspect of tool-use isn't really part of this draft.

   The second author is a non-native English speaker who participates in
   IETF standardization work across several working groups.  The
   following describes current practice in detail, as a basis for the
   discussion that follows.

3.1.  Monitoring and Triage

   Incoming mailing list traffic is large and often spans multiple
   simultaneous threads.  An AI assistant is used to scan the mailbox
   and identify threads or messages that appear relevant, including
   those that may warrant a reply.  The author then reads the original
   messages directly.  The AI provides a signal; the reading and
   judgment are the author's.

3.2.  Forming a Position

   After reading, the author thinks about the issue independently.  This
   step is not delegated.  The AI may be used at this stage to stress-
   test an argument --- to articulate the strongest counter-position, or
   to explore whether an alternative interpretation holds --- but it
   does not originate the position.  The author decides what to think
   before asking AI to help express it.

Farrell & Feng           Expires 7 February 2027                [Page 5]
Internet-Draft            LLM Email Discussions              August 2026

3.3.  Drafting in English

   Once the author has decided what to say, an AI assistant produces an
   English draft.  English is not the author's first language, and
   producing precise, idiomatic technical prose in a second language
   carries a real cognitive cost.  AI removes that cost without changing
   who is responsible for the ideas.

   The author reviews this draft critically --- not for grammatical
   correctness, but for fidelity.  If the output is too long, too
   polished, too neutral, or does not accurately represent the intended
   position, revisions are requested.  This can take several rounds.
   The test is not "does this read well" but "does this say what I
   meant."

   One specific issue encountered is that AI output tends toward an
   artificially "balanced" stance --- hedging between positions instead
   of committing to one.  Draft outputs may under-commit when compared
   to the author's actual position.  This effect may not show up as a
   "tell" visible to readers of the eventual message, but can be visible
   to the author as one.

3.4.  Final Review and Send

   The author personally reviews the final text before sending.  The AI
   does not send mail autonomously.  The author takes full
   responsibility for anything sent under their name.

3.5.  The Role of Odyssey

   The author has developed a personal AI agent called Odyssey
   (https://github.com/meetodyssey).  Odyssey maintains long-term
   context across conversations --- what subjects the author cares
   about, how they normally reason, what positions they have taken over
   time.  The long-term goal is for AI-assisted output to become
   increasingly consistent with how the author would write
   independently, as the system accumulates a genuine model of the
   author's thinking rather than producing generic fluent prose.  This
   points toward something important: the right relationship between a
   person and their AI tools is one that deepens over time, becoming
   more accurate to the individual rather than more generic.

4.  Human and AI: Complementary Capabilities

   This section is also written by the second author.

Farrell & Feng           Expires 7 February 2027                [Page 6]
Internet-Draft            LLM Email Discussions              August 2026

   The issues described above reflect a genuine tension.  To resolve it,
   it helps to be precise about what AI systems actually do well and
   what they do not.

4.1.  What AI Does Well

   AI language systems operate effectively within known boundaries.
   Given a well-defined problem space --- an established body of
   knowledge, a clear communication goal, a defined set of constraints
   --- AI can draft, translate, and refine text with speed and
   consistency no individual can match; identify relevant prior work
   across large corpora; stress-test arguments by generating counter-
   positions; and execute repetitive cognitive tasks without fatigue.

   For participants in international technical communities, the language
   function alone is significant.  The ability to express a precise
   technical idea in idiomatic English is not the same as having the
   idea.  AI collapses the gap between the two, allowing non-native
   speakers to participate on more equal terms.

   More fundamentally, AI excels at operating within accumulated
   knowledge.  The existing literature of a field, its terminology, its
   conventions, its prior decisions --- this is exactly the kind of
   material AI systems are built to handle.

4.2.  What AI Does Poorly

   AI systems have a structural limitation that is frequently
   underestimated: they cannot originate.

   They recombine, extrapolate, and interpolate within the space of what
   they have been trained on.  This is not a temporary limitation
   awaiting a better model.  It is a consequence of what these systems
   are.  Genuine innovation --- the creation of new conceptual territory
   rather than more efficient mapping of existing terrain --- remains a
   human capacity.

   The ideas that change a field do not emerge from pattern completion.
   They emerge from the collision of lived experience, accumulated
   frustration, specific domain knowledge, and the kind of lateral
   connection that has no prior example to learn from.  The recognition
   that the current framework is wrong, or that the question being asked
   is the wrong question, is not available to a system trained to
   operate within existing frameworks.

Farrell & Feng           Expires 7 February 2027                [Page 7]
Internet-Draft            LLM Email Discussions              August 2026

4.3.  The Symmetry

   Humans have the inverse limitation.  We are slow to acquire and
   integrate knowledge within established boundaries.  Learning a field
   takes years.  Staying current across adjacent areas is nearly
   impossible for any individual.  Expressing ideas precisely in a
   second language imposes a constant cognitive cost.

   AI removes these constraints.  A researcher with AI assistance can
   engage with a much larger body of prior work, express ideas more
   precisely across language barriers, and iterate on arguments more
   rapidly than was previously possible.

   The symmetry is clean: humans originate, AI executes.  Humans open
   new territory; AI operates efficiently within it.  Neither is
   complete without the other.

4.4.  A Collaborative Paradigm

   From this symmetry, a working paradigm emerges.

4.4.1.  The Core Principle

   Humans originate.  AI executes.

   The position, the argument, and the judgment are formed by the human
   before AI involvement begins.  AI is used to express, refine,
   translate, or stress-test what the human has already worked out.  The
   human reviews AI output not for grammatical correctness but for
   fidelity to their actual position.  The human takes full
   responsibility for anything sent or published.

   This boundary is not always clean in practice.  Using AI to stress-
   test an argument can surface considerations the human had not thought
   of, which then reshape the position.  This is legitimate --- the AI
   is functioning as a thinking partner within a bounded space, not as
   an originator.  The human remains the decision-maker about what to
   accept and what to discard.  What falls outside this paradigm is
   delegating the thinking itself: asking AI what position to take, what
   arguments to make, or what conclusions to draw --- and signing the
   output.

Farrell & Feng           Expires 7 February 2027                [Page 8]
Internet-Draft            LLM Email Discussions              August 2026

4.4.2.  Transparency as a Norm

   The workflow described in Section 3 was questioned.  The author
   described it in detail.  This exchange --- uncomfortable at first ---
   produced this document.  Transparency is not a concession to critics
   of AI assistance.  It is what makes the collaboration legitimate.  An
   author who can describe exactly how AI was used, and who can stand
   behind the resulting text as an accurate representation of their
   position, has nothing to hide.  An author who cannot answer those
   questions has a different problem, and it is not the AI.

4.4.3.  The Deepening Relationship

   Generic AI assistance produces generic-sounding output.  A system
   that accumulates genuine knowledge of an author's thinking,
   positions, and style produces output that more accurately represents
   them --- not because it is deceiving anyone, but because it has
   become a better instrument.

   This is the direction Odyssey points toward.  Over time, the gap
   between "what the author would have written" and "what the AI-
   assisted author sent" narrows.  The tool becomes more personal, more
   accurate, and paradoxically more transparent: the output is more
   genuinely the author's, not less.

4.4.4.  What Becomes Possible

   The significance of this paradigm is not merely defensive --- not
   simply a justification for a practice that would otherwise be
   suspect.  It is expansive.

   Things that were previously impossible for one person to accomplish
   --- engaging seriously with a large technical field while also
   pushing its boundaries, participating in international discourse
   while thinking in another language, tracking developments across
   multiple working groups while developing original contributions ---
   become achievable.

   This is the door the current moment opens.  Not AI replacing human
   contribution, but AI extending the reach of what any human can
   contribute.  The combination of human originality and AI execution
   capacity creates possibilities that neither possesses alone.

5.  Risks

   The risks of AI-assisted writing are real, but frequently
   misdescribed.  Attempting to describe them precisely should help.
   Both authors contributed to this section.

Farrell & Feng           Expires 7 February 2027                [Page 9]
Internet-Draft            LLM Email Discussions              August 2026

   *  AI replacement of thought.  A primary risk is not AI assistance
      but the delegation of thinking itself --- asking AI what to
      believe, not just how to express a belief.  This produces output
      that is fluent but not genuine, and over time it degrades the
      author's own capacity for independent thought, as well as putting
      the reader in an impossible position.

   *  Drift from the author's position.  An author may form a genuine
      position but accept an AI draft that misrepresents it --- more
      confident, more agreeable, or more hedged than intended ---
      without noticing.  Careful review of AI output exists to catch
      this.  It requires the author to know their own position well
      enough to recognize when it has been distorted.

   *  Reader inability to distinguish.  AI-assisted expression and AI-
      replaced thinking may produce similar surface output.  Readers
      cannot easily tell them apart.  This erodes the trust that makes
      mailing list discussion valuable, and it creates an asymmetry that
      disadvantages responsible users alongside irresponsible ones.

   *  Homogenization of discourse.  AI systems have characteristic
      tendencies --- toward confidence, toward agreement, toward certain
      rhetorical patterns.  Widely adopted without discipline, these
      tendencies flatten the diversity of perspective that technical
      communities depend on.  A list where everyone's prose sounds
      similar, however polished, is a less productive list.

   *  Writing what one does not believe.  Distinct from the above, an
      author may knowingly send AI output that does not reflect their
      actual position, using the tool as a shield against
      accountability.

   *  Emitting gibberish.  AI tooling is imperfect, if we end up with
      multiple senders using AI tools and so essentially have AI tools
      running a substantive discussion, we are more likely to end up
      with gibberish.

   *  Discussion based on bad information.  AI tooling might emit text
      that is based on outdated information or even hallucinated
      material.  Senders need to check messages they send, and may need
      to be very familiar with the topic(s) being discussed, to ensure
      this does not occur.

   While not properly described as a risk, the two authors of this draft
   do not currently agree as to whether it would be an overall positive
   or negative were there to be no "tells" visible in messages emitted
   with the assistance of AI tooling.  More discussion is needed on
   that:-)

Farrell & Feng           Expires 7 February 2027               [Page 10]
Internet-Draft            LLM Email Discussions              August 2026

6.  Initial Discussions

   This draft was raised on the IETF "discuss" list with some discussion
   ensuing. [ldref] This section aims to record points raised in a way
   that may be more easily found than the list archive, as well as a few
   points raised off-list.  Thus far, there has been no discussion
   solely on the github repo for this draft.

6.1.  Possibly Relevant Policies Elsewhere

   Some other relevant policies and discussions were brought to the
   authors' attention and could feed into discussion of an IETF policy:

   *  one from W3C [w3cpol], which seems very relevant to this
      discussion, but that also seems nearly as tentative as this draft

   *  the EU AI act might contain some relevant clauses [euaiact50] that
      might (or might not) call for disclosure of use of AI tooling in
      contexts such as standards-development

   *  the Irish courts have a very recent policy on the use of LLMs in
      court documents [iecourt]

   *  the IRTF's RASPRG discussed related topics at IETF-126 [rasprg]

6.2.  Other points

   *  one poster expressed that using LLMs during discussions seemed
      less "honest" to them, whereas using LLMs for Internet-draft text,
      tool development or testing/analysis seemed more acceptable

   *  the IETF may have some "self-defense" mechansisms that help us
      avoid some of the worst potential problems that could arise from
      using LLMs in discussions, e.g., physical (or online synchronous)
      meetings and slow progress making it easier to spot LLM usage over
      extended durations

   *  another risk is that people using LLMs might deliberately
      manipulate LLM tools to produce outputs that aren't really
      intended as part an IETF discussion, but rather to try to distort
      or disturb that discussion

   *  attempts to ban the use of LLMs as described here won't work

   *  any policy in this space won't be known to new participants who
      are therefore more likely to not conform to that policy (while
      this could be argued generally about any policy, it's perhaps very
      relevant here, given current trends)

Farrell & Feng           Expires 7 February 2027               [Page 11]
Internet-Draft            LLM Email Discussions              August 2026

   *  the IETF could encourage doing better in this space rather than
      simply denounce such uses

   *  there's a risk that if we don't try tackle this issue IETF mailing
      lists might end up like [moltbook]

   *  a sender's use of LLMs may make it harder for a reader to
      distinguish between a sender that is less well informed but who
      will learn from discussion, versus a sender that is not actually
      willing to learn from a discussion (and who perhaps doesn't
      understand that the LLM output is gibberish)

   *  over time, readers learn to expect and better interpret senders
      who use their own "voice" - interposing an LLM risks changes to
      that in ways that remove a tool IETF discussants have used for
      decades, (and for senders, LLM version changes might totally
      change the "voice" that readers perceive)

   *  some people just skip message they consider "vacuous and wordy"
      and believe many of those are LLM outputs

   *  some consider the "voice" of known participants as helpful in
      evaluating their inputs

   *  messages largely based on LLM outputs may be "boring"

   *  LLMs might be helpfule in discussions about the history of draft
      and e.g. whether or not merging some drafts might be good or bad

   *  "unfinished" LLM generated outputs seemingly describing something
      may waste the time of the (possibly many) readers of a message

   *  some code-of-conduct for use of LLMs might be useful, along with
      bans for breaches

   *  to the extent we assume "good faith" participation, use of LLMs
      may change our threat model for participation

   *  assuming that IETF participants have the resources and/or time to
      engage with LLM tools could be yet another barrier to
      participation

   *  IETF discussions should be for humans, and respect the amount of
      time readers have available to consider messages

   *  one could use LLM tools to shorten the messages one sends

Farrell & Feng           Expires 7 February 2027               [Page 12]
Internet-Draft            LLM Email Discussions              August 2026

   *  senders are 100% responsible for what they send - use any tooling
      (or send any message) at the risk of your reputation

   *  we are capable of bikeshedding on the name of a mailing list for
      disucssion of this topic, but we should have such a list

   *  perhaps the end result of this discussion should be a wiki page
      that evolves and not a policy expressed in an RFC

   *  well-connected people, with good English language skiils, have had
      an easier time in the past, maybe these tools might change that

   *  the IETF should actively take a position and declare requirements
      (for discussion) that suit our needs

   *  requiring declaration of LLM tool use might be like cookie-banners
      and become so common as to not be useful

   *  so far, someone's LLM output concluded "the discussion is
      constructive rather than polarised" ;-)

   *  LLM tools can help with translation, but increased acceleration
      (in terms of producing output) may inevitably correlate with a
      lack of understanding from the sender

   *  LLM use may disrupt reader's evaluation of sender reputation over
      time

   *  one poster suggested publishing community-specific guidance that
      participants could feed to their AI agents (e.g., bottom line up
      front, avoiding walls of jargon, double-checking assertions before
      posting) so that AI-assisted contributions start out closer to
      community norms

7.  Recommendations

   These are extremely tentative recommendations, that may be wrong, but
   that seem worth considering:

   *  Form your position before engaging AI.  The argument should be
      yours before the draft exists.

   *  Review AI output for fidelity, not just correctness.  If the AI
      has softened, generalized, or shifted your position, correct it
      before sending.

   *  Always be transparent.  Describing your workflow in detail builds
      trust rather than eroding it.

Farrell & Feng           Expires 7 February 2027               [Page 13]
Internet-Draft            LLM Email Discussions              August 2026

   *  Question senders if you think they are using AI tooling as to what
      they are doing.  Doing that on-list should be considered
      acceptable, if it is not done in an accusatory manner.

   Less tentatively, the IETF should develop guidelines for use of AI
   tooling when sending messages (esp email) in IETF discussions.  That
   won't be easy and will be a moving target, but absent such guidance,
   confidence in email discussions may evaporate, which would cause
   significant damage to the IETF.

8.  Conclusion

   It's too early to say really.

9.  IANA Considerations

   This document makes no request of IANA.

10.  Security Considerations

   Mischievous IETF participants could include AI prompts inside
   messages used in IETF discussions that could form part of an attack
   on participants who use AI tooling.  Such text could, for example,
   only be present in the text/html part of a multipart/alternative
   email and so might not be rendered in a presentation of a mailing
   list archive.  Presumably AI tool users will need to mitigate such
   threats in any case, so the new aspect here is perhaps only the use
   of IETF archives as the distribution medium for AI prompt attacks.

   Otherwise, see the section on risks.

11.  Acknowledgments

   The second author used https://github.com/meetodyssey in preparing
   and discussing the above text.

   The first author made no use of AI tooling.

12.  References

12.1.  Normative References

   [RFC5378]  Bradner, S., Ed. and J. Contreras, Ed., "Rights
              Contributors Provide to the IETF Trust", BCP 78, RFC 5378,
              DOI 10.17487/RFC5378, November 2008,
              <https://www.rfc-editor.org/info/rfc5378>.

Farrell & Feng           Expires 7 February 2027               [Page 14]
Internet-Draft            LLM Email Discussions              August 2026

   [RFC8179]  Bradner, S. and J. Contreras, "Intellectual Property
              Rights in IETF Technology", BCP 79, RFC 8179,
              DOI 10.17487/RFC8179, May 2017,
              <https://www.rfc-editor.org/info/rfc8179>.

12.2.  Informative References

   [euaiact50]
              "EU AI Act Explorer", n.d., <https://ai-act-service-
              desk.ec.europa.eu/en/ai-act/article-50>.

   [iecourt]  "Practice Direction on the Responsible Use of Generative
              Artificial Intelligence in Court Documents", July 2026,
              <https://www.courts.ie/practice-directions/full-practice-
              direction?url=practice-direction-on-the-responsible-use-
              of-generative-artificial-intelligence-in-court-documents>.

   [ldref]    "Dealing with LLMs in IETF discussions draft", July 2026,
              <https://mailarchive.ietf.org/arch/msg/
              ietf/3VaBJ6pEdhtkpOtnZYA_HcVHejU/>.

   [moltbook] "Moltbook wikipedia page", August 2026,
              <https://en.wikipedia.org/wiki/Moltbook>.

   [rasprg]   "RASPRG IETF-126 meeting", July 2026,
              <https://datatracker.ietf.org/meeting/126/session/rasprg>.

   [w3cpol]   "Use of Large Language Models in Standards Work", March
              2026, <https://www.w3.org/TR/2026/NOTE-llms-standards-
              20260324/>.

Appendix A.  Change Log

A.1.  Draft-00

   *  This is based on email and github interactions between the
      authors.

A.2.  Draft-01

   *  Reflect points raised on the IETF "discuss" list.

Authors' Addresses

Farrell & Feng           Expires 7 February 2027               [Page 15]
Internet-Draft            LLM Email Discussions              August 2026

   Stephen Farrell
   Trinity College Dublin
   College Green
   Dublin
   Ireland
   Email: stephen.farrell@cs.tcd.ie

   Chong Feng
   Email: fengchongllly@gmail.com

Farrell & Feng           Expires 7 February 2027               [Page 16]