CCSAagent-levellinggroup-standardkey
Published 2 March 2026, effective 1 June, under TC1/WG1 (Internet applications and AI), project 2025-CCSA-002. The standard defines a capability classification method for AI agent systems and the technical requirements per level, for use by both providers and consumers when grading agent systems; the English title itself says "AI agent system". Thirty drafting organisations including China Mobile and CAICT as leads, all three operators, Baidu, BAAI, Huawei, Ant, Alibaba, ZTE and iFLYTEK. Why it matters for the agent gateway: capability levelling is the base for later assessment and certification, and paired with the Internet Society trusted-agent assessment specification (open for comment until 1 October) it forms the two pillars of "levelling plus assessment" on the group-standard side. Source: CCSA official standard detail page (first-party).
CCSATC3agent-networkproject-approvalkey
During the 48th TC3 plenary (31 March to 2 April, Wuhan), the WG3 SWG2 agent-network subgroup held its first meeting; over ten agent projects were initiated, including "Agent network — IP networking research for agent tasks" led by FiberHome and the "Agent-network terminology" industry standard jointly with China Telecom, within 55 plenary approvals spanning agent networks, computing-power networks and digital-twin networks; WG1 additionally approved the "IP bearer network architecture and key technologies for 6G" study (FiberHome with China Unicom). The subgroup third meeting is already on the association calendar for 22 September. Background: China Unicom publicly pushed for the subgroup within TC3 in late 2025. Why it matters: domestic agent-network standardization now has a standing organizational home, rolling out terminology, IP networking and task bearing item by item — directly comparable with your RTGWG draft line. Source: FiberFire official WeChat via NetEase (trade media) and the CCSA meeting calendar.
CCSAA2A-TAN-L4NetMCPkey
In his keynote "Agent interconnection, cloud-network-compute coordination", academician Wu Hequan argued agent workloads are pushing the Internet from "people operating" to "machine autonomous loops", with long context, high concurrency and 24/7 operation, and that both compute and network architectures must evolve toward agent-native design. The conference issued the joint pledge to accelerate AN L4 deployment, whose third direction is to start the next-phase standardization of the multi-agent collaboration target architecture and the A2A-T carrier agent interconnection protocol. China Mobile Research VP Duan Xiaodong laid out three innovations (new access: the agent communication network ACN; new protocol: the interconnection base stack; new application), and released the industry first carrier-network agent — the Bixing agent platform with the NetMCP protocol. Observation: A2A-T has moved from open source (OpenAN) onto the domestic standardization agenda, while Bixing plus NetMCP is another carrier-built gateway protocol stack instance parallel to the AONP line. Source: IT168 (16 April 2026, trade media).
ITU-TSG13agent-standardizationkey
The ITU-T Study Group 13 plenary (future networks and emerging technologies; designated lead study group on AI by WTSA-24) closed with 52 new work items started and 36 standards completed; the 176-member Chinese delegation submitted 194 contributions (64 percent) and leads 36 new items (69 percent). On AI agents specifically, standardization advanced systematically: terminology definitions, network architecture, gateway functions and trust assessment all have international standard projects, with China leading the field. The first IMT-2030 network terminology Recommendation was also approved, and multiple FG AINN (AI-native networking focus group) pre-study results were converted into standard projects. Observation: "gateway functions" and "trust assessment" hit the agent gateway squarely — the ITU-T gateway architecture line, the domestic assessment system and the IETF gateway drafts are advancing in parallel. Source: Communications World Weekly (authoritative trade media).
ITU-TACNYSTR-NACY-aiasfirst-party
The official ITU-T work programme lists YSTR.NAC-roadmap, "Networking for AI agent collaboration — Standardization status and roadmap", a technical report targeted for completion in November 2026, supported by China Telecom, CAICT, China Unicom, BUPT and Zhejiang Lab; the companion YSTR.NAC defines the framework of networking for AI agent collaboration in future networks; YSTR.ACN defines the AI-agent communication network in IMT-2020 networks and beyond. Q17/13 additionally runs Y.aias-reqts (AI Agent as a Service requirements, driven by ETRI). JCA-AI, the cross-SG AI coordination group, lists "AI Agents and Multi-Agent Systems" among its 2026-2027 focus areas. Why it matters: the ITU-T ACN shares both name and substance with the ACN concept China Mobile promoted — heterogeneous agent identity, discovery and collaboration networking now have a canonical ITU-T vehicle. Source: ITU-T official work programme page (first-party).
3GPP6GTR-22870requirementsfirst-party
SA1 study FS_6G_REQ (TR 22.870, 6G use cases and service requirements) was approved at TSG SA#111 in March with 200+ use cases, explicitly including the "6G AI agent collaboration with third-party AI" use case family and requirements such as exposing services to authorised third-party AI agents; the report frames AI agents as automated intelligent entities capable of intent understanding, contextual reasoning, self-learning and collaborative decision-making within the 6G system. Rel-20 now carries 13 six-G studies (requirements, architecture, radio, application enablement, security, media, charging, management), with TR 38.914 (6G radio scenarios) approved in June. Why it matters for the agent gateway: 3GPP wrote agents into the Stage-1 requirements as first-class citizens of the 6G system, so later network-side specifications for exposure, authorization and collaboration all have a requirements anchor — carrying agent traffic in carrier networks moved from an idea to a requirement. Source: 3GPP Release-20 page (first-party).
3GPPSA6application-enablement6G
SA6 established the study on 6G application enablement (FS_6G_APP), taking the SA1 TR 22.870 use cases (including third-party AI agent collaboration and exposing services to authorised third-party AI agents) as input to derive application-enabler level requirements, aligned with the IMT-2030 vision and industry initiatives such as GSMA Open Gateway and ETSI OpenCAPIF. Work plan: SA6#68 (25-29 August) proposals and initial analysis, moderated discussions in September and October, SA6#70 (17-21 November) to finalise the SID for SA plenary approval; the output TR 23.xxx will form the basis for normative 6G application enablement work. Observation: the application-enablement layer is exactly where agent-gateway capabilities can sink into and be exposed by the network — if the GSMA Open Gateway API体系 converges with agent collaboration needs, carrier APIs gain a new class of agent-enabler calls. Source: 3GPP Release-20 page (first-party); SID draft S6-255552 readable via browser.
3GPPSA26G-architectureagent-identity
SA2 study FS_6G_ARC (architecture for the 6G system) is about 50 percent complete, targeted for March 2027; topics at the late-August Prague week included AI-agent discovery and communication, AI-agent identity and authorization, the characteristics of generative-AI traffic patterns, support for AI model training and inference, and exposure of AI services through 6G core network functions. The report notes none of these design choices existed in 5G, because 5G was not designed with AI as a first-class network citizen. Why it matters for the agent gateway: the architecture group is deciding boundary questions such as whether agents get network-native identity, and whether discovery and communication ride network functions or the application layer — the answers will shape the interface between agent gateways and the 6G core, making this the best window for network-side researchers to weigh in. Source: Tech Times (28 August 2026, trade media; agenda items from its 3GPP meeting tracking).
6 items Daily Brief · 2026-09-06
wechatagent-gatewayexplainer
WeChat piece explaining the agent gateway as the connective layer that discovers available agents, routes messages, checks permissions, invokes tools and records run state. Useful plain-language explainer of exactly the control loop a gateway owns. Source: WeChat public account of the China Institute of Communications (permalink).
productAI-gatewayopen-source
On 2 September the open-source 1Panel project opened its AI gateway to the community, providing standalone AI-gateway deployment via official images, free for teams of up to ten. Another signal that gateway capability is commoditising into self-hosted tooling. Source: 1Panel official site (2026-09-03).
ITU-TSG13agent-network
New SG13 contribution converting FG AINN outcomes into proposed Annex II text on AI agents within networks. Part of the SG13 agent line alongside YSTR.ACN and YSTR.NAC. Source: ITU-T SG13 contribution #934 (first-party).
ITU-TSG13agent-network
Proposed technical report YSTR.SO-ACN on the secure operation of the AI-agent communication network (ACN) — the canonical ITU-T vehicle for agent-native 6G networking, sharing name and substance with the ACN concept China Mobile promotes. Secure operation is squarely agent-gateway territory. Source: ITU-T SG13 contribution #935 (first-party).
ITU-TSG13agent-network
New SG13 TR proposal on potential requirements and a functional framework for networking that supports AI agent collaboration, complementing YSTR.NAC and its roadmap. Source: ITU-T SG13 contribution #945 (first-party).
ITU-TSG13agent-network
New SG13 cloud-computing work item proposal (framework and functional architecture), providing the compute and cloud perspective for the SG13 agent-collaboration networking series. Source: ITU-T SG13 contribution #956 (first-party).
This brief was AI-assisted in collection and summarization, then human-reviewed before publication.