Telecom Giants Launch Open AI Initiative
The GSMA has launched Open Telco AI, a new collaborative initiative backed by major industry players. The goal is to accelerate the development of open-source, interoperable AI solutions specifically designed for the reliability and security needs of telecom networks.
The initiative was launched to address the significant performance gap of general-purpose AI models in the telecommunications sector. Tests have shown that generic large language models can provide 30-40% incorrect responses to specific technical telecom queries and have been found to invent non-existent 5G spectrum policies. This performance issue is a critical barrier to innovation, with only 16% of generative AI deployments in the telecom industry being applied to network operations. The new initiative aims to solve this by creating specialized benchmarks, datasets, and models tailored for the unique needs of telecommunications. AT&T and AMD are founding supporters of the project. AT&T is contributing a family of open telco-models trained on publicly available data, which are designed to be hardware and cloud-agnostic. AMD, along with its cloud partner TensorWave, is providing the necessary computing power for model training, fine-tuning, and evaluation through its GPU platforms. The collaboration extends to more than two dozen organizations worldwide. Key contributing partners include China Mobile, Deutsche Telekom, Google Cloud, IBM, KDDI, Nvidia, Orange, SK Telecom, Telefónica, and Vodafone, among others. The initiative will establish a definitive open-source hub for telco-specific AI resources. A "Telco Capability Index" will be used to measure how well different AI models perform on a range of telecom-specific tasks. The project's scope also includes a focus on energy efficiency, safety, and multilingual support. The goal is to ensure that the AI models are not only accurate but also sustainable and trustworthy for deployment in critical network infrastructure.