On September 14, 2026, at the opening ceremony of China's National Cybersecurity Awareness Week in Jinan, Shandong, authorities released the results of the 2026 AI-empowered cybersecurity application testing exercise. It was run by the National Computer Network Emergency Response Technical Team/Coordination Center (CNCERT) together with the Institute of Information Engineering at the Chinese Academy of Sciences, the National Information Technology Security Research Center and Shandong Radio and Television, and guided by roughly sixteen bodies including CAC's Cybersecurity Coordination Bureau, the Supreme People's Court, and the cyber and technology departments of MIIT, the Ministry of Public Security, the central bank and the customs administration. The testing set eight scenarios and opened registration to the public: 151 organisations and 251 teams took part, drawn from cybersecurity, information and communications, AI, finance and broadcasting firms as well as universities and research institutes. Huawei supplied the Ascend AI hardware and software environment. Thirty-two teams recorded strong results across scenarios that included AI-driven attack defence, source-code vulnerability discovery, network-traffic threat detection, alert-log noise reduction, large-model safety-guardrail testing, AIGC-generated-image detection, detection of malicious AI-agent behaviour, and anomaly detection for IPTV accounts.
Our takeNotice that one scenario is detecting malicious behaviour by AI agents (scene 7), sitting right beside large-model guardrail testing. That is a state acknowledging autonomous agents as a threat surface, not just a tool. The whole exercise also ran on a single domestic chip stack - a real evaluation, but also an ecosystem play worth reading with that framing in mind.