Global Telcos Launch 'Open Telco AI'

The initiative directly addresses the reality that general-purpose AI models struggle with the telecom industry's specific needs. Tests show that even advanced AIs provide incorrect responses 30-40% of the time on technical telecom topics and can "hallucinate" non-existent standards, creating significant regulatory risks. AT&T and AMD are anchor partners in this effort. AT&T is contributing a family of open-source AI models trained specifically on public telecom data, while AMD, along with its partner TensorWave, is providing the essential high-performance GPU computing power needed for training and fine-tuning these complex models. A broad coalition of over two dozen organizations, including operators like China Mobile, Deutsche Telekom, Orange, and Vodafone, and vendors such as Nvidia and Google Cloud, have thrown their support behind the project. This wide-ranging collaboration aims to pool data, models, and expertise from industry and academia. The project establishes a new portal for shared resources and a "Telco Capability Index" to benchmark AI model performance on network-specific tasks. This addresses a key industry challenge, as currently only 16% of generative AI deployments in the telecom sector are used for network operations. This focus on specialized AI is critical as the industry moves toward 6G. Future networks will rely on AI for real-time network optimization, predictive maintenance, enhanced security, and managing complex applications like autonomous vehicle fleets and holographic communication. The initiative aims to overcome major hurdles in AI adoption, such as the shortage of talent with combined AI and telecom expertise, and the difficulty of integrating new AI systems with complex legacy network infrastructure. By creating open, accessible, and specialized AI tools, the GSMA and its partners aim to lower the barrier to entry for operators and accelerate a move toward more intelligent, automated networks.