Mei Ri Jing Ji Xin Wen·2025-07-29 14:00Core Viewpoint - China Telecom is fully embracing artificial intelligence (AI) and upgrading its corporate strategy from "cloud transformation" to "intelligent cloud transformation" to enhance digital transformation across various industries [3][4]. Group 1: AI Integration and Services - China Telecom aims to create an integrated intelligent cloud service system combining computing power, platforms, data, models, and applications to empower digital transformation in various sectors [3]. - The company is enhancing its AI safety governance capabilities by developing a comprehensive AI dynamic protection system [3]. - The "Xingchen Textile Intelligent Body" showcased at WAIC 2025 can intelligently adjust process parameters and inspect fabric quality during production, achieving a 99% on-time delivery rate [4][6]. Group 2: Performance Improvements - China Telecom has improved the efficiency of its reasoning engine and matched computing power with minimal latency through integrated computing and networking services, resulting in over a 40% increase in computing resource utilization and a 30% reduction in average reasoning latency [4]. - The company plans to launch terminal-as-a-service and intelligent agent-as-a-service systems to further promote AI application and empower industrial intelligent transformation [4]. Group 3: Industry Applications and Models - China Telecom has developed over 80 industry-specific large models and more than 20 intelligent agent applications targeting sectors such as industry, emergency response, and education [5]. - The company emphasizes the need for an end-to-end security assurance system that encompasses infrastructure, data, models, and applications to address the diverse safety risks associated with AI [5]. Group 4: Security Measures - To tackle security challenges in the AI era, China Telecom proposes a four-in-one security model comprising early warning, assessment, protection, and operation [5]. - The "Jianwei" security model has achieved a 98% accuracy rate in fault alert identification, and the foundational security measures for large models will be open-sourced domestically to enhance resistance against AI attacks [5].