A new industry challenge is urging telecom operators, AI researchers, and startups to build large language models (LLMs) capable of identifying and explaining the root causes of network faults — a persistent, costly problem for operators. Launched today, the AI Telco Troubleshooting Challenge is supported by GSMA, ETSI, IEEE GenAINet, ITU, and TM Forum, bringing together diverse organisations aiming to advance AI-driven network operations.
The competition includes several tracks, each targeting a specific aspect of network troubleshooting. One track evaluates how effectively an LLM handles unfamiliar or previously unseen faults. Another focuses on small, efficient models designed to run at the network edge where compute and power are constrained. A third assesses an AI system’s ability to present clear, traceable reasoning when diagnosing issues. Additional categories address edge-cloud security and the creation of AI services for application developers.
Organisers view the challenge as a practical way to surface solutions that help operators restore service faster and automate more operational workflows. Entries will be judged on accuracy, speed, interpretability, and security. Alongside the main community partners, the initiative has support from Huawei, InterDigital, NextGCloud, RelationalAI, xFlowResearch, and technical advisers from AT&T.
Industry groups outline expectations for AI in network operations
ETSI highlights that many challenges operators face today stem from long-standing gaps in model generalisation and the limitations of edge computing. Dario Sabella, Chair of ETSI MEC, said the challenge provides teams with specialised datasets and infrastructure that can “accelerate the adoption of Telco AI.” He emphasised that compact, efficient language models at the edge could make AI deployments more accessible across the telecom industry.
IEEE GenAINet sees the challenge as an opportunity to validate ideas that are central to building more autonomous networks. Prof. Merouane Debbah, General Chair of IEEE GenAINet ETI, said the event draws attention to critical topics such as unseen fault detection, interpretability, and edge-efficient AI — areas he described as essential for “making AI-native telecom infrastructures a reality.”
The ITU views the initiative as part of its broader work to reduce barriers for innovators experimenting with telecom-focused AI systems. Seizo Onoe, Director of the ITU Telecommunication Standardisation Bureau, said the organisation’s global challenges are designed to give developers access to computing resources, datasets, and mentorship needed to help their projects achieve meaningful impact.
This initiative builds on earlier efforts to benchmark AI models for telecom tasks. Curated datasets such as TeleLogs and tools developed by the GSMA Open-Telco LLM Benchmarks community have provided standardised ways to measure model performance. Those efforts, including a public leaderboard for telco-specific scenarios, have helped operators and researchers understand how LLMs perform on tasks like fault diagnosis, trend analysis, and reasoning under uncertainty.
GSMA notes that the industry’s focus on improved reasoning and diagnosis is driven by the scale of financial losses caused by network faults each year. Louis Powell, Director of AI Technologies at GSMA, said root cause analysis “is an important pain point for operators,” and better solutions could enhance reliability while reducing ongoing operational costs.
The challenge aims to accelerate the development of models that combine robust reasoning with operational efficiency and scalability.
AT&T, serving as a technical adviser to the challenge, recently ran experiments that illustrate the potential of smaller AI models. The company fine-tuned a 4-billion-parameter model that outperformed other systems on the TeleLogs root cause analysis benchmark, including larger frontier models. Andy Markus, Chief Data Officer at AT&T, described the challenge as bringing together “an important business problem and a technical opportunity,” and said increased industry collaboration can extend this progress.
Organisers emphasise that the challenge provides teams with access to resources that are often difficult to secure independently, creating a space to test new ideas. As networks grow more complex, they expect more operators to adopt AI systems that can diagnose issues rapidly, explain their reasoning, and operate efficiently across edge and cloud environments.
For many partners involved, the AI Telco Troubleshooting Challenge represents a practical step toward telecom networks that are more reliable, adaptive, and easier to manage at scale.
(Photo by Google DeepMind)
Want to discover how IoT is transforming telecoms and connectivity? Attend the IoT Tech Expo in Amsterdam, California, or London to explore how innovations in 5G, edge computing, and IoT are shaping the future of networks and services. The event is part of TechEx and co-located with other leading technology conferences.
Telecoms News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars on the TechForge events page.