Research Challenges in Coupling Artificial Intelligence and Network Management
draft-irtf-nmrg-ai-challenges-06
Note: This ballot was opened for revision 06 and is now closed.
Ballot question: "Is this draft ready for publication in the IRTF stream?"
Brian Trammell
Yes
Comment
(2026-08-18)
Sent
Thank you for addressing my review comments!
Colin Perkins
Yes
Comment
(2026-08-17)
Not sent
I made a detailed technical review of -04 and the authors made an update that addressed my comments. The later changes further improved the draft. I'm happy for it to be published.
David Oran
Yes
Comment
(2026-08-17)
Sent
Nice and comprehensive, though not terribly easy to extract the important take-aways given the volume of material. TLDR won't easily get a reader the most important things. Despite that, I think it's ready to publish. One minor comment relates to section 6 on network data. Recent research has highlighted the dangers of poisoned measurement data leading A.I.- based network management actions far astray, for example: https://arxiv.org/abs/2508.06394. I think this deserves mentioning.
Alvaro Retana
No Objection
Comment
(2026-08-25)
Sent
§9.2/§9.3 scope boundary: The distinction between adversarial AI (§9.2) and unsafe or incorrect AI outputs (§9.3) could be clarified. An adversary manipulating an AI component may ultimately cause the same operational harm discussed in §9.3, making the boundary somewhat unclear. I suggest explicitly stating that *general* adversarial AI/ML research is out of scope, while security threats to AI components that affect network-management security, safety, or correctness remain in scope. This would make the intended scope and relationship between the two sections clearer.
Ryo Yanagida
No Objection