The "Cyberspace Security Situation Analysis Report (2025)" jointly released by the National Computer Virus Emergency Response Center, 360 Technology Group Co., Ltd. and others shows that from June 2024 to July 2025, a total of 59 artificial intelligence-related security incidents occurred at home and abroad. In the future, we need to be alert to the risks of artificial intelligence abuse such as deep forgery fraud, misuse of artificial intelligence to carry out cyber attacks, and artificial intelligence infringement.

This reporter learned this from the third Cyberspace Security (Tianjin) Forum held on the 23rd.

Du Zhenhua, a senior engineer at the National Computer Virus Emergency Response Center, said that artificial intelligence technology will achieve significant breakthroughs in 2025. The cost of training large models continues to decrease, and artificial intelligence applications are rapidly evolving from generative AI to intelligent AI. The "double-edged sword" effect of technology has become increasingly obvious. The abuse of generative AI has led to the "real-time scene-based" upgrade of deep forgery, making it more difficult to identify video face-changing fraud.

Artificial intelligence security risks are mainly divided into network security risks, data security risks and model security risks. According to incomplete statistics, compared with 2024, artificial intelligence-related network and data security incidents and risks will increase significantly in 2025, of which network attacks account for 29%, data security incidents account for 26%, and security hazard incidents account for 25%.

Data provided by the National Computer Virus Emergency Response Center. Photo provided by interviewee

The Cyberspace Security (Tianjin) Forum has been held twice. This forum focuses on the security risk challenges faced by the accelerated development of the information and innovation industry, the dynamic balance of citizen rights protection and digital economic development, and other cutting-edge and hot topics in cyberspace security. (Reporters Yang Wen and Huang Jianglin)