T/BMISC 001-2024 Data Security Specifications for the Application of Large-scale Models in the Medical Field

Scope

Main technical content: Medical data is an extremely important type of data, which includes sensitive information such as patient medical records, imaging data, physiological parameters, and genetic data. It not only concerns personal privacy but also involves social public interests and national security. Recently, there have been a series of breakthroughs in the research of large-scale models, with parameter sizes reaching the level of hundreds of billions, and they have extremely strong capabilities in text generation and understanding. While large-scale models are widely used, they also give rise to a series of privacy protection and data security issues. In the medical field, large-scale models have great application prospects in improving medical diagnosis, drug research and development, and clinical decision-making. However, due to the fact that the training and application of large-scale models involve massive medical data, the risks of personal privacy leakage, ethical issues, and discrimination and bias caused by their improper use require widespread attention. This specification applies to suppliers or manufacturing enterprises with the ability to develop medical large-scale models, and can guide medical institutions in deploying medical large-scale model products/solutions. It can also serve as a basis for medical institutions to select or evaluate the safe application of large-scale models, aiming to regulate the behavior of data application of large-scale models in the medical field, clarify the corresponding data privacy protection measures and model safety evaluation requirements, ensure that the privacy and security of medical data are properly protected, and promote the technological innovation and development of large-scale models in the medical field.

Development Information

Related Standards