Standards bodies treat AI and networking in two directions: using machine learning to run mobile networks better, and designing networks that carry and host AI workloads. The 3GPP study on AI for the 5G New Radio air interface, part of Release 18, covered three use cases (channel state information feedback, beam management, and positioning), while the International Telecommunication Union lists "artificial intelligence and communication" as one of six usage scenarios for 6G, formally called IMT-2030.
Key takeaways
- 3GPP Release 18 studied AI and machine learning for the radio interface as a study, not as a finished feature.
- A separate 3GPP work item looked at AI inside the radio access network, building on a study of energy saving, load balancing, and mobility optimization.
- ITU Recommendation ITU-R M.2160 lists six 6G usage scenarios, one of them artificial intelligence and communication.
- ITU expects candidate 6G radio technologies to be submitted between February 2027 and February 2029, so 6G AI features are plans, not products.
Two meanings of "AI and networking"
The phrase covers two different engineering questions. The first is AI for the network: a base station or core network function uses a trained model to predict something and adjust its own behavior. The second is the network for AI: the radio system is designed so that data collection, model training, and inference can be spread across phones, base stations, and edge computers. Standards documents keep these apart, and so does the table below. The table is a simplified orientation built from the sources listed at the end, and it describes planning documents, not deployed products.
| Direction | Body and document | What it covers |
|---|---|---|
| AI for the radio link | 3GPP Release 18 study on the NR air interface (TR 38.843) | Channel state information feedback, beam management, and positioning |
| AI inside the access network | 3GPP Release 18 work item for NG-RAN, following study TR 37.817 | Data collection and signaling for energy saving, load balancing, and mobility optimization |
| AI in network management | 3GPP SA5 study on AI and machine learning management | Management capabilities and services for AI and machine learning in the 5G system |
| Network as a platform for AI | 3GPP SA2 study on 5G system support for AI-based services; ITU-R M.2160 scenario | Carrying AI traffic and supporting distributed AI applications |
What 3GPP studied in Release 18
3GPP is the partnership of standards organizations that writes the specifications for mobile systems. A 3GPP overview article by Juan Montojo, the rapporteur of the radio-layer study, describes the Release 18 work as a study running through the whole release, tracked under the code FS_NR_AIML_air. It looked at whether AI could improve performance or cut signaling overhead in three areas. For channel state information, the part of the link where a phone reports radio conditions to the base station, frequency-domain compression was the agreed direction at the time of writing, and time-domain prediction was still under consideration. For beam management, spatial and temporal prediction looked promising. For positioning, the study covered both direct approaches, such as fingerprinting, and AI-assisted ones.
The study also set out a general framework: stages of an AI algorithm, levels of collaboration between the phone and the base station, how a model's life cycle is managed, and how datasets are built. In most use cases the model was assumed to run at either the base station or the phone. Only channel state information was expected to explore a two-sided model, in which both ends cooperate closely, and the article described that choice as still under discussion.
Inside the radio access network
A second thread concerned the next-generation radio access network, known as NG-RAN. A 3GPP highlights article dated August 8, 2022 lists a work item on AI and machine learning for NG-RAN, coded NR_AIML_NGRAN, and says it followed a study documented in TR 37.817. That study examined data collection and use cases in energy saving, load balancing, and mobility optimization. In plain terms, these are tasks where an operator wants to switch idle equipment off, spread users across cells, or hand a moving user to the best cell, and where learned predictions might beat fixed rules.
The same article names two further efforts: an SA5 study on how to manage AI and machine learning functions inside the 5G system, and an SA2 study on using the 5G system as a platform for AI-based services. Both are studies of possibilities.
What ITU says about 6G
The ITU's Radiocommunication Sector sets the global framework for mobile generations. Its IMT-2030 page states that the framework recommendation was adopted by Study Group 5 on September 28, 2023, and approved by the Radiocommunication Assembly in November 2023. A December 1, 2023 ITU press release says the assembly met in Dubai from November 13 to 17 and that the framework identifies 15 capabilities, nine of them derived from existing 5G systems.
The page lists six usage scenarios: immersive communication, hyper reliable and low latency communication, massive communication, ubiquitous connectivity, artificial intelligence and communication, and integrated sensing and communication. The press release summarizes the AI scenario as communications networks supporting AI-powered applications. The page itself gives only the expansion of the abbreviation, so detailed technical requirements for that scenario must be read in the recommendation text, which could not be opened for this article.
Timeline and limits of what is settled
The ITU page says candidate radio interface technologies are to be received between February 2027 and February 2029, and the press release says final 6G standards are expected to be approved by 2030. On the 3GPP side, the articles cited here date from 2022, so they describe plans and early study results. Later outcomes, including which Release 18 and later features reached specifications, commercial devices, or live networks, were not confirmed from primary sources for this article and are left out. A reader evaluating a vendor claim about AI-native networks can compare it with the dated, document-level descriptions above.
Questions to ask about an AI networking claim
- Does the claim concern a 3GPP study, a finished specification, or a shipping product?
- Does the model run on the phone, at the base station, or in the core network?
- Which task is improved: radio link feedback, beam selection, positioning, energy saving, load balancing, or mobility?
- Is the gain measured on a public benchmark, in a simulation, or in a live network?
FAQ
Is AI part of 5G standards today?
Release 18 included studies and work items on AI and machine learning in the radio interface, the radio access network, and network management. The sources used here describe the plans from 2022 and do not confirm the final contents of each specification.
What does the ITU mean by artificial intelligence and communication?
It is one of six IMT-2030 usage scenarios. ITU's December 2023 press release describes it as communications networks supporting AI-powered applications.
When will 6G exist?
ITU expects candidate technologies to be submitted from February 2027 to February 2029 and final standards to be approved by 2030. Commercial availability depends on later 3GPP specifications and operator decisions.
Does AI in a network mean a chatbot?
No. In standards documents, AI usually means a trained model that predicts a quantity such as channel state, beam direction, or traffic load, and the network uses the prediction to adjust its own settings.
Related reading
- How AI changes everyday communication
- AI systems that build other AI systems
- DeepSeek explained
- AI in delivery and transport
More in AI Tools and Industry.
Last verified: October 2026
Sources
- IMT-2030 (Connecting the world and beyond), ITU Radiocommunication Sector, 2026
- ITU advances the development of IMT-2030 for 6G mobile technologies, ITU press release, December 1, 2023
- AI/ML for the NR air interface, Juan Montojo, 3GPP Highlights, November 2022
- Finding AI in 3GPP, 3GPP, August 2022