With the deep integration of artificial intelligence (AI) into public security work, public security organs across China have actively explored the application of large AI model technologies to develop intelligent infrastructure including training datasets, knowledge bases, and public security-specific large models, injecting technological momentum into the continuous enhancement of the new-type combat capabilities of public security organs. However, in the process of improving policing efficiency, issues such as data security breaches, algorithmic bias, and adversarial attacks have occurred frequently. The problem of black-box algorithms has aroused widespread public concern, and ensuring the safety, reliability, and controllability of AI has become a prominent issue, which is increasingly threatening national security. This study systematically analyzes various security risks in AI-driven public security services, and proposes responsive strategies to strengthen confidential security management and data security protection from the perspectives of ideological risks, policy and legal norms, technical construction risks, and regulatory governance mechanisms. It provides insights for public security organs at all levels to seize the opportunities of AI empowerment while using AI technologies in a safe manner.
Lu Yulei (Wed,) studied this question.