[Technical] The claim that “TS 23.501 includes 5QI 88 for AI/ML traffic” is questionable/misleading: 5QI values are standardized for specific service characteristics, but “AI/ML traffic” is not a single service type; the contribution should cite the exact clause and standardized service description for 5QI 88 (and avoid implying it generically covers AI/ML).

[Technical] The paper asserts “traffic characteristics input is the most urgent topic” but provides no concrete SA4 deliverable impact (e.g., which SA4 specs/TRs would be blocked without it, what parameters are missing, and what normative outputs are expected), making the urgency argument weak and hard to action in ToR.

[Technical] The proposed SA4 focus on “media service-related AI traffic characteristics” is under-defined: it does not specify whether this means conversational media with AI processing, AI-generated media, model/feature streaming, sensor-to-media fusion, or inference offload—each has different latency/jitter/burstiness and mapping to 5QI/GBR behavior.

[Technical] “Multi-modal AI representations and formats” is not clearly within SA4’s traditional remit unless tied to concrete media formats/codecs or RTP/ISOBMFF carriage; otherwise it risks overlapping with SA2 service architecture or external SDOs, so the ToR text needs sharper boundaries and interfaces.

[Technical] The document references updating TR 26.927 and “the 6G study” but does not identify the target work item, release, or responsible group for “6G study,” creating ambiguity on where SA4 would actually capture the proposed updates.

[Technical] The argument that “most AI traffic (text/application data) is outside SA4 scope” is broadly reasonable, but the paper overreaches by categorically excluding MCP/A2A/agent protocols without analyzing cases where such traffic is synchronized with media sessions (e.g., real-time captions, prompts controlling media generation), which could create gaps if ToR wording becomes too restrictive.

[Technical] The “AI codec vs. AI model distinction” is valid, but the contribution does not define what an “AI codec” means in 3GPP terms (bitstream syntax? interchange format? model compression? feature coding?), risking inconsistent interpretation and scope creep in ToR.

[Technical] The recommendation to exclude “general AI/ML quality evaluation” is sensible, yet the paper does not acknowledge existing SA4 quality frameworks (e.g., subjective/objective media QoE, speech/audio quality, immersive media) that could be extended to AI-generated media; a blanket exclusion could unintentionally block needed evaluation for AI-based media codecs.

[Technical] The “Embodied AI has the most stringent real-time requirements” statement is asserted without any quantitative characterization (latency budget, reliability, jitter, uplink/downlink symmetry), making it a weak basis for ToR changes and potentially misdirecting SA4 priorities.

[Technical] The claim that “current mobile networks meet most requirements of today’s AI services (chatbots)” is not substantiated and ignores emerging real-time multimodal assistants (audio/video) that may already stress uplink latency and jitter; this weakens the rationale for anchoring ToR changes primarily on 6G.

[Editorial] The contribution repeatedly references an “attachment” containing suggested ToR updates, but the provided text does not include the actual proposed wording; without explicit tracked changes or proposed ToR text, SA4 cannot evaluate scope, deliverables, or overlaps.

[Editorial] Several terms are introduced without definition or 3GPP references (“multi-modal AI,” “MCP/A2A,” “agent protocols,” “AI-related formats”), which is problematic for a ToR discussion where precise scope language is essential.

[Editorial] The paper cites “SA1 and SA2 in Release 18” work but provides no specific document identifiers/clauses beyond the 5QI example; stronger referencing is needed to justify that “general AI traffic characteristics” are already sufficiently covered elsewhere.

[Editorial] The narrative mixes ToR-scoping arguments with speculative technology positioning (e.g., 2030/6G anchoring) without clearly separating what is proposed as ToR text versus background opinion, reducing clarity on what SA4 is being asked to decide.