[Technical] The proposed scope for WT#3a includes “Media Client/AF interaction with network-managed AI agents” and “network-provided AI inferencing/training,” but it does not state any SA4-specific assumptions (e.g., whether these AI agents/services are in CN, edge, or AF domain), risking overlap/contradiction with SA2 KI#18/KI#19 boundaries (core-embedded AI vs. AI entities outside the core).

[Technical] The introduction summarizes KI#18 and KI#19 solution variants but does not map them to concrete media use cases or media KPIs (latency, jitter, throughput, QoE, synchronization), so it is unclear why the new WT#3a text belongs under clause 6.4 (media aspects) rather than remaining purely architectural.

[Technical] The text asserts KI#19 is “more directly relevant to SA4 media work” yet does not justify this with any explicit media-related mechanisms (e.g., media session control, codec adaptation, streaming policy, XR media pipelines), making the relevance claim weak and potentially contentious.

[Technical] The scope item “intent/goal-based media service interactions” is underspecified: it does not indicate whether intent is between UE–AF, AF–network exposure, or UE–network, nor how it relates to existing 3GPP exposure frameworks (e.g., NEF/CAPIF concepts), which could lead to inconsistent requirements later.

[Technical] “Media functions as exposed capabilities/tools for AI agents” lacks clarity on what constitutes a “media function” in SA4 terms (codec, renderer, compositor, analytics, QoE estimator, etc.) and whether exposure is via AF APIs, network exposure, or application-layer protocols, leaving the scope too open-ended for a study item.

[Technical] The proposal references KI#7 (Network Exposure) and KI#22 (6G Computing Support) as secondary but does not explain their specific linkage to media AI workflows (e.g., edge compute placement for inference, exposure of media QoS/QoE telemetry), missing an opportunity to anchor WT#3a to identifiable SA4 deliverables.

[Technical] The handling of KI#19.5 (“AI traffic characteristics”) is left as an editor’s note “reserved pending SA4 input,” but the new text does not define what SA4 is expected to contribute (traffic models, QoS/QFI mapping, buffering tolerance, uplink vs downlink asymmetry), risking an unactionable placeholder.

[Editorial] The contribution adds multiple editor’s notes indicating incompleteness; for TR text, excessive “TBD/pending” language in newly introduced clauses can reduce readability and may be better captured as explicit study questions or a work plan rather than embedded notes.

[Technical] The summary states UE AI agents are out of scope for KI#18, but the WT#3a scope includes Media Client interaction with AI agents; without explicitly reconciling this, readers may misinterpret whether UE-resident AI for media is in-scope for SA4 under WT#3a.

[Editorial] Clause numbering is potentially inconsistent: adding 6.4.1.1, 6.4.1.3, and 6.4.1.4 implies a missing 6.4.1.2; if intentional, it should be explained or aligned to existing TR 26.870 structure to avoid broken references.

[Editorial] The text heavily paraphrases SA2 TR 23.801-01 key issues; it should include precise citations (clause numbers) for KI#18/KI#19 and solution variants to avoid ambiguity and ease cross-WG review.

[Technical] The proposed scope does not mention security/privacy or governance aspects of AI agents interacting with media functions (authorization, exposure control, data minimization), which are likely critical when “media functions” and user media/QoE data are exposed to AI entities.

[Editorial] Terminology such as “agentic protocols,” “AI agents,” and “tools” is introduced without definition; TR 26.870 should either define these terms or reference an agreed SA2/SA1 terminology source to prevent divergent interpretations across WGs.