Abstract: In the context of the deep implementation of the national "Artificial Intelligence+" strategic action in the railway industry, it is of great significance to study the safety technology of large models in the railway industry. This paper studied the security policies of artificial intelligence large models at home and abroad, analyzed the security requirements of large models in training, deployment, and operation stages based on the current development status of railway industry large models, and studied the security protection technologies for the entire life cycle of railway industry large models, including infrastructure security, dataset construction security, large model construction security, and large model application security. The aim is to comprehensively enhance the security protection capabilities of railway industry large models and provide security technology solutions for railway professional research and application of large models.