Hi Martin, I am replying to your note, but commenting more generally.
First, the claim has been made by several people that a document belongs to the working group in the IETF. I agree with this point 100%. In fact, so much so that it might make sense for the IETF to reconsider the boiler plate used for RFCs to bring home that point. That's not for us to do here, tho. I dispute the notion that in the vast overwhelming majority of cases these documents undergo radical revision. In my experience, the vast overwhelming majority of changes are actually proposed by the authors themselves. From time to time, someone else proposes a change, and even in those cases, in my experience, the vast overwhelming majority of the changes that are accepted are done with the approval of the author. The remainder of cases are, in my experience, rare. And even rarer is the day when an author has to be removed for not approving such changes.
Why do I go into such details? Because while the policy must allow for the case where authors are removed, the far more common case should set reader expectations. This document should align to that point.
Second, regarding AI, I understand the point that Jay is making, and it does raise concerns. However, I agree with Brian that at an academic honesty demands that we not take *credit* for other people's work. The RFC series is rooted in academia and we should respect that attribute. *Moreover, *if someone finds that an AI agent has been misbehaving in some way, having at least some understanding of the provenance of the work becomes important.
I agree with your major point, however, that the authors must take responsibility for their work. In my view, this is an important enough point to even recommend that a question confirming that be placed on the RPC intake form, and even reaffirmed during AUTH48. It's even important to understand this question earlier, and here perhaps streams might differ by degrees, but that's not an RFC problem.
By the way, I am not looking forward to the day when "Agentic AI" signs up for datatracker accounts and starts participating in our activities. It *will* happen, as it did with USENET all those many years ago in the net.suicide, where Mark V. Shaney held prominence.[1]*. Just imagine what rough consensus will look like THEN.
Eliot* Say what you will about Rob Pike, but he has a wicked and imaginative sense of humor.
[1] https://en.wikipedia.org/wiki/Mark_V._Shaney On 04.06.2026 06:22, Martin Thomson wrote:
On Thu, Jun 4, 2026, at 12:21, Brian E Carpenter wrote:* If, however, a substantial part of the document was created by AI this must be disclosed, typically in the Acknowledgements section. This requirement is to avoid any confusion about the authorship of the document and to ensure that its readers are not misled.Why let people duck responsibility that way? Every use of AI disclaimers I've ever seen was used in an attempt to avoid taking responsibility for the text. That can be OK in some contexts, but if we are talking about contributions to RFCs, this sort of requirement only shields people from owning up to the words they are asking the community to publish. I'd prefer if there be a *prohibition* on such notices. That might also justify prohibiting this acknowledgment: This document was prepared using 2-Word-v2.0.template.dot. As a reader of an RFC, I almost negatively care about the tooling in the general sense. There are places where tooling is important context, but those would be exceptional circumstances, not the common case (I've mentioned tooling in RFC 8448, for example, as it is relevant in that case).I definitely don't think the RSWG should get into a discussion of copyright,I don't think we can avoid it entirely, but it seems sensible to keep it out of scope for this effort.
OpenPGP_0x87B66B46D9D27A33.asc
Description: OpenPGP public key
OpenPGP_signature.asc
Description: OpenPGP digital signature
-- rswg mailing list -- [email protected] To unsubscribe send an email to [email protected]
