AI
Etisalat UAE and Khalifa University Push AI Into the Architecture of 6G

Etisalat UAE and Khalifa University have published peer-reviewed research proposing an AI-native architecture for future 6G networks.The design introduces a dedicated AI layer capable of learning, predicting faults and helping networks operate with greater autonomy.
[For more news, click here]
Mobile technology generations have traditionally been measured by what they can carry: more data, lower latency, and more connected devices. 6G is starting to raise a different question. What if the network could understand what was happening and decide what to do next?
That is the ambition behind new peer-reviewed research from Etisalat UAE and Khalifa University, which proposes an AI-native architecture that embeds artificial intelligence in the network’s fundamental design rather than adding it later as a management tool.
Published in the IEEE Open Journal of the Communications Society, the research develops an architecture that the two organizations first outlined earlier this year. It adds a dedicated “AI-plane” alongside a telecom network’s traditional user, control and management layers. The goal is to give future networks the ability to continuously learn, reason, predict and act.
AI Moves From Add-On to Network Layer
Mobile networks have been using AI in areas such as traffic forecasting, energy optimization, anomaly detection, and network planning. However the systems were not originally designed around AI.
Under the Etisalat UAE and Khalifa University framework, AI becomes a first-class network layer. The AI-plane would coordinate intelligence across radio networks, the core network, edge infrastructure and cloud systems, allowing AI agents distributed across the network to share information and make decisions.
An AI-assisted network uses intelligence to improve individual functions. An AI-native network assumes intelligence is required for the network to operate as designed. The research identifies six dimensions separating those two models, including how AI is embedded into the architecture and how much responsibility remains with human operators.
A Network That Can Run More of Itself
The researchers describe autonomous connectivity as a long-term ambition. Rather than having engineers manually configure every aspect of the network, AI systems could increasingly take on planning, service assurance, optimization, and recovery. Human operators would still define objectives, policies, and limits, while the network determines more of the operational steps needed to meet them.
Using the TM Forum’s six-level autonomy model, the researchers target levels 3 to 4 for validated network domains, with levels 4 to 5 envisioned for routine operations over the longer term. That would mark a significant shift in telecom operations, and the network would not simply report that performance had deteriorated. It could potentially identify the problem, predict its effect, choose an intervention and make the adjustment before customers noticed.
Digital Twins Provide the Safety Layer
Giving AI more control over a live telecom network raises an obvious concern: operators cannot let experimental models learn from mistakes on infrastructure serving millions of customers. The proposed architecture addresses this with a network digital twin.
The digital twin serves as a continuously synchronized replica of the live network, providing operators with a controlled environment to train, test, and evaluate new AI models and policies before deployment.
Models could also run in “shadow mode,” making decisions based on live network conditions without controlling production systems. Engineers can then compare the AI’s decisions with what actually happened. This creates a path toward greater autonomy without giving up control right away. For telecom operators, this could become one of the most important elements of AI-native 6G.
The UAE Is Building More Than One Piece of the 6G Stack
The research is part of a broader UAE effort to shape 6G technology. Earlier this month, Tech Revolt reported on Khalifa University’s RFGPT research, which is testing whether foundation models can understand radio-frequency signals directly rather than relying entirely on specialist engineers to interpret the wireless spectrum.
That work addresses the intelligence needed to understand the radio environment. The latest research addresses how to integrate that intelligence into the wider network.
A separate October study from Etisalat UAE, Khalifa University and NYU Abu Dhabi focused on another layer: how future operators could use centimetre-wave spectrum to deliver roughly three times the capacity of existing 5G networks while retaining comparable coverage using existing macro sites.
Taken together, the projects cover three important questions. How should 6G understand the wireless environment? How should operators physically build the network? And how should intelligence be distributed across it?
The commercial 6G market is still years away. Etisalat UAE and Khalifa University’s roadmap places research and prototyping in 2026–2027, standardization and pre-commercial trials in 2027–2029, and broader commercial deployment from 2029 through 2032 and beyond.
The architecture of 6G is still being debated across organizations including ITU IMT-2030, 3GPP, O-RAN Alliance, ETSI and the AI-RAN Alliance.
Countries and companies that contribute research during this phase can potentially influence the technical standards adopted internationally. The race toward 6G may ultimately be less about which country launches the first commercial service and more about who helps define the architecture everyone else eventually uses.
Related Articles
UAE Researchers Are Teaching AI to Read the Wireless Spectrum
Etisalat UAE and DataRobot Push Sovereign Agentic AI Into Enterprise Workflows
Gulf Service Providers Expect AI Network Services to Drive New Revenue, Ciena Survey Finds





