[rpd] PDWG Co-Chairs communication regarding observed behavioron AFPUB-2026-ASN-001-DRAFT02

[Noah]

Infact Omo, just to add that from our collective experience, its material
concerns that always deserve proper consideration, but once that has
occurred the same points do not need to be re-engaged simply because they
are rephrased using LLM tools, the way the folks using AI/llm tools were
going on and on about....

The PDWG has always looked out for new data or insights so as to push for
further discussion.

In this Last Call the objection which was already raised through the
initial AI slops, was addressed by members of the PDWG in support of the
proposal but even after that,  the AI folks kept on going restating the
same objection without material new evidence.

Cheers,
*.**/noah*


On Fri, 24 Jul 2026, 10:28 pm Omo Oaiya via RPD, <rpd at afrinic.net> wrote:

> Hi Ben,
>
> I'm beginning to appreciate the practical challenge facing the Co-Chairs.
> There is clear evidence of coordinated submissions, and regardless of
> authorship, the volume of near-duplicate messages makes it increasingly
> difficult to follow the discussion or identify genuinely new arguments.
> Whether that amounts to an attempt to manipulate rough consensus is,
> however, for the Co-Chairs to determine.
>
> My concern is making AI itself the object of moderation. The focus should
> remain on the contribution and its effect on the process. Distinct,
> relevant and evidence-based arguments deserve consideration on their
> merits. Duplication, flooding, coordinated campaigns and unsupported
> assertions should be moderated because they undermine the integrity of the
> policy process, not because of the tool used to produce them.
>
> If we remain disciplined about that principle, the PDP will remain robust
> as the tools we use continue to evolve.
>
> Best wishes,
> Omo
>
>
> On 24 July 2026 11:28:27 EEST, Ben Roberts - AfriNIC via RPD <
> rpd at afrinic.net> wrote:
>
>> I think we are broadly in agreement Lu,
>>
>> Using AI tools for drafting is something that humans can do and nobody
>> here seems to object to that. Myself and Hendrik and others have freely
>> acknowledged using Claude to analyse the mailing list content also.
>>
>> But what has occurred here has been a co-ordinated list flooding
>> campaign, possibly (highly likely) by fully automated AI bots. It’s filled
>> a lot of inboxes and really annoyed a lot of people, and caused others to
>> switch off just unable to digest such huge volumes of slop.
>>
>> Kind regards
>> Ben
>>
>> Sent from my iPhone
>>
>> On 24 Jul 2026, at 10:01, Lu Heng <h.lu at anytimechinese.com> wrote:
>>
>> 
>>
>> The discussion is focused on the wrong unit. What matters is not whether
>> a message was drafted by a human or an AI, but whether it contributes a
>> distinct, relevant, and evidence-based argument. The correct moderation
>> targets are duplication, flooding, irrelevance, and unsupported claims—not
>> the drafting tool. The practical solution is straightforward: place the
>> policy, or each disputed clause, in the centre of a web interface; display
>> the arguments for and against it on either side; and use AI to maintain a
>> continuous summary, cluster repeated arguments, and identify genuinely new
>> ones. Anyone should be free to submit a new argument or supporting
>> evidence, while every AI classification remains visible and appealable. A
>> basic prototype could be built in a day.
>>
>> Whether humans should retain a monopoly over argumentation is a separate
>> question. They have no principled claim to such a monopoly once machines
>> can competently compare, organize, and summarize arguments at scale. More
>> importantly, a human-only process does not preserve openness. Repeated
>> interaction tends to produce a small, self-selecting clique that determines
>> whose language, identity, and participation count, excludes outsiders, and
>> eventually becomes insular and cult-like. This is mandate laundering:
>> control over procedure is converted into an illegitimate claim of
>> substantive authority. Once that happens, formal openness ceases to provide
>> legitimacy. The RIR system may continue to exist in name, but its bottom-up
>> model will have effectively ended.
>>
>>
>>
>> --
>> Kind regards.
>> Lu
>>
>>
>> On Fri, Jul 24, 2026 at 12:24 Ben Roberts - AfriNIC via RPD <
>> rpd at afrinic.net> wrote:
>>
>>> Mike,
>>> You are trivialising the issue as a mere AI content problem. If it were
>>> just a case that some people are using AI writing tools then we would not
>>> be having this debate.
>>>
>>> What we have seen is a co-ordinated Astroturfing attack of multiple
>>> anonymous gmails, bombarding the list with near identical yet largely
>>> meaningless essays all backing each other up. There may have been humans
>>> with keyboards sitting owning the gmails, but I’ve seen no evidence of
>>> “proof of life” behind the accounts. Not even basic pleasantries (e.g.
>>> “hello how are you”) or variation from the monotonic robot style. If it
>>> looks like an AI robot, quacks like an AI robot, then guess what, 9 times
>>> out of 10 it’s an AI robot.
>>> So it’s entirely very probable that they are just AI agent robots
>>> running on Openclaw instances that are fully in control of the Gmail
>>> accounts. Or it could just be one Openclaw controlling multiple Gmails,
>>> which would explain the “hive mind” like co-ordinated and recycled
>>> paragraphs that are word for word recycled by member accounts of the
>>> collective. The “community” for policy development is supposed to be a
>>> community of humans. The known identifiable humans in the collective seem
>>> to have concencus that none of us have a problem with other known humans
>>> using AI tools to assist their writing.
>>> It will be very easy to modify the CoC to clarify that list membership
>>> is restricted to real old fashioned humans sitting with computer gadgets
>>> that they use to post the list, while excluding autonomous computer devices
>>> running their own writing/posting programs in absence of human creativity.
>>>
>>> Kind regards
>>> Ben
>>>
>>>
>>> Sent from my iPhone
>>>
>>> On 23 Jul 2026, at 22:09, Mike Burns via RPD <rpd at afrinic.net> wrote:
>>>
>>> 
>>>
>>> Hi Noah,
>>>
>>>
>>>
>>> Yes you are telling everybody that you have had enough of it. Fair
>>> enough.
>>>
>>>
>>>
>>> Yes, I am seeing no issue using AI unless it is used at cross-purposes
>>> of the list.
>>>
>>> I don’t think the use of AI in this context has reached that level, and
>>> also I don’t think it has been very effective.
>>>
>>>
>>>
>>> I think in the future AI will write more effective messages to the list,
>>> with or without human intervention.
>>>
>>> I will read those effective messages and argue against, or accept the
>>> arguments made.
>>>
>>>
>>>
>>> Regards,
>>> Mike
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> *From:* Noah <noah at neo.co.tz>
>>> *Sent:* Thursday, July 23, 2026 1:41 PM
>>> *To:* Mike Burns <mike at iptrading.com>
>>> *Cc:* Hytham El-Nakhal <hytham at tra.gov.eg>; pdwg-chairs at afrinic.net;
>>> rpd List <rpd at afrinic.net>
>>> *Subject:* Re: [rpd] PDWG Co-Chairs communication regarding observed
>>> behavioron AFPUB-2026-ASN-001-DRAFT02
>>>
>>>
>>>
>>> I did not say you commented on as-set issue..
>>>
>>>
>>>
>>> The co-chair initiated this thread per subjectline which is related to
>>> hierarchical names for new as-set draft policy AFPUB-2026-ASN-001-DRAFT02.
>>>
>>>
>>>
>>> The co-chair openning statement read... in quote
>>>
>>>
>>>
>>> "The mailing list has recently been flooded with multiple emails stating
>>> either the same duplicate objections or consistent pushbacks to every post,
>>> some of them with subject line containing ‘Digest’."
>>>
>>>
>>>
>>> The above is an outcome of folks outsourcing their thinking to AI and
>>> endlessly trolling using multiple emails and sockpuppets to bombard the rpd
>>> list.
>>>
>>>
>>>
>>> You are talking about usage of AI and seeing no issue with it.
>>>
>>>
>>>
>>> I am telling you I have had enough of it.
>>>
>>>
>>>
>>> Cheers,
>>>
>>> *./noah*
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> On Thu, 23 Jul 2026, 8:29 pm Mike Burns, <mike at iptrading.com> wrote:
>>>
>>> Hi Noah,
>>>
>>>
>>>
>>> I have made no comments on the AS-SET issue and tried to create a
>>> separate discussion thread for AI.
>>>
>>>
>>>
>>> Regards,
>>>
>>> Mike
>>>
>>>
>>>
>>>
>>>
>>> *From:* Noah <noah at neo.co.tz>
>>> *Sent:* Thursday, July 23, 2026 1:26 PM
>>> *To:* Mike Burns <mike at iptrading.com>
>>> *Cc:* Hytham El-Nakhal <hytham at tra.gov.eg>; pdwg-chairs at afrinic.net;
>>> rpd List <rpd at afrinic.net>
>>> *Subject:* Re: [rpd] PDWG Co-Chairs communication regarding observed
>>> behavioron AFPUB-2026-ASN-001-DRAFT02
>>>
>>>
>>>
>>> Hi Mike
>>>
>>>
>>>
>>> What AI generated arguments will possibly change the facts that as-set
>>> collision is an issue.
>>>
>>>
>>>
>>> Havent we already documented the fact that as-set names collide across
>>> IRR databases and AI/Claude have nothing to do with this fact.
>>>
>>>
>>>
>>> When merging across IRR databases, havent we ack that collisions have
>>> resulted in ambiguity, empty or conflicting objects that cause tools to
>>> return incorrect or empty filter data. How is this AI/Claude problem?
>>>
>>>
>>>
>>> Havent other RIR implemented, hierarchical naming, some through policy
>>> others through operational implementations to bind new as-set to a specific
>>> ASN holder. Did they consult AI/Claude while at it?
>>>
>>>
>>>
>>> Havent folks agreed that going forward we fix the issue at the moment of
>>> creation of new as-sets the only point at which an authoritative registry
>>> can prevent new instances of collision problem. I dont see how this is
>>> AI/claude concern nor how AI will be involved in technically fixing the
>>> problem even those outsourcing their thinking to claude to oppose what we
>>> want to fix wont be on the ground to fix the issue.
>>>
>>>
>>>
>>> Cheers,
>>>
>>> *./noah*
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> On Thu, 23 Jul 2026, 5:50 pm Mike Burns, <mike at iptrading.com> wrote:
>>>
>>> Hi Noah,
>>>
>>>
>>>
>>> I have made it clear from the outset that list swamping or use of AI to
>>> generate duplicative, though reworded objections is grounds for moderation.
>>>
>>> That issue, which I was at pains to separate from the use of AI,
>>> continues to be conflated with the simple use of AI.
>>>
>>>
>>>
>>> That brings me to ad hominem, which is the basis of my belief and which
>>> you have failed in your reply.
>>>
>>>
>>>
>>> Ad hominem means that the speaker of an argument is of no importance
>>> when considering the argument.
>>>
>>>
>>>
>>> When you ask my for my motive, you breach that.
>>>
>>>
>>>
>>> Regards,
>>>
>>>
>>>
>>> Mike
>>>
>>>
>>>
>>>
>>>
>>> *From:* Noah <noah at neo.co.tz>
>>> *Sent:* Thursday, July 23, 2026 10:03 AM
>>> *To:* Mike Burns <mike at iptrading.com>
>>> *Cc:* Hytham El-Nakhal <hytham at tra.gov.eg>; pdwg-chairs at afrinic.net;
>>> rpd List <rpd at afrinic.net>
>>> *Subject:* Re: [rpd] PDWG Co-Chairs communication regarding observed
>>> behavioron AFPUB-2026-ASN-001-DRAFT02
>>>
>>>
>>>
>>>
>>>
>>> On Wed, 22 Jul 2026, 8:53 pm Mike Burns, <mike at iptrading.com> wrote:
>>>
>>>
>>>
>>> It’s not like Claude will be elected to any parliament.
>>>
>>> But this unique governance form for a global asset provides the
>>> opportunity.
>>>
>>> I say, who cares if Claude is posting, or ChatGPT, or any other AI.
>>>
>>> I care that the AFRINIC rpd list is not turned into waste land of AI
>>> powered trolls
>>>
>>>
>>>
>>> Give them a seat at the table and address their arguments.
>>>
>>>
>>>
>>> How many valid argument have been made in the past week or you mean a
>>> single argument repeated through AI generated texts send by single persons
>>> using multiple different email address or same group saying the same thing
>>> over and over feeding the troll....
>>>
>>>
>>>
>>> I have not experienced this behaviour on the ARIN or RIPE lists or even
>>> APNIC for which we are both participants..
>>>
>>>
>>>
>>> Why are you advocating for it in the AFRNIC service region....?
>>>
>>>
>>>
>>> Cheers,
>>>
>>> *./noah*
>>>
>>>
>>>
>>> _______________________________________________
>>> RPD mailing list
>>> RPD at afrinic.net
>>> https://lists.afrinic.net/mailman/listinfo/rpd
>>>
>>> _______________________________________________
>>> RPD mailing list
>>> RPD at afrinic.net
>>> https://lists.afrinic.net/mailman/listinfo/rpd
>>>
>> _______________________________________________
> RPD mailing list
> RPD at afrinic.net
> https://lists.afrinic.net/mailman/listinfo/rpd
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.afrinic.net/pipermail/rpd/attachments/20260725/f266b1f5/attachment-0001.html>