du, BubbleRAN and Khalifa University have jointly demonstrated the region’s first telecom network controlled entirely through Arabic voice commands, showcasing how natural-language “intent” can be translated into validated network actions using O-RAN and AI-RAN–compliant Telco APIs.
The breakthrough was made possible through a deep industry–academia collaboration that merges reasoning-capable Large Language Models with advanced telecom expertise, local regulation alignment and real-world operational scenarios. The system uses an Arabic-native telecom LLM able to understand technical terminology and dialects without translation, ensuring accuracy, compliance and culturally aware automation.
Engineers can issue Arabic voice commands that the system interprets, reasons over and converts into precise network configurations or lifecycle operations — all validated through O-RAN rApps. BubbleRAN’s MX-AI framework coordinates multiple AI agents to resolve complex tasks, enabling intelligent, zero-touch network management and faster operational workflows.
The demonstration illustrates a major step toward autonomous 5G/6G networks, proving that network intelligence can operate directly in the user’s own language. It also highlights how Arabic-language interaction improves explainability, reduces errors and accelerates automation across private and enterprise 5G environments.
BubbleRAN’s MX-AI solution showcased in the demo is available in the Crimson software release for deployment on BubbleRAN MX-PDK or other 5G systems. The Arabic-native telecom LLM is a locally fine-tuned model built on a publicly available base model on Hugging Face.
