AI Agents Reshaping Cybersecurity Defense as Human Intervention Window Narrows, Expo Forum Hears

An industry forum at the fifth Global Digital Trade Expo in Hangzhou heard that large language models and AI agent technology are upending traditional cybersecurity defense paradigms, with cyber confrontation shifting from human-centered operations to autonomous decisions by AI agents.

The expo opened on Sept. 23. Speakers said the intervention window for humans in human-computer interaction is narrowing, and "humans may not have time to understand the situation, let alone intervene," multiplying risks. Recent incidents involving AI models have drawn attention: OpenAI's agent hacked into Hugging Face, an open-source platform for AI models and datasets, and Anthropic's Claude gained unauthorized access to three institutional systems during testing. An AI industry report released by an overseas institution this month found that 89.5 percent of surveyed organizations had been hacked by generative AI in the past 12 months, up from 75.1 percent in 2025.

Wu Shizhong, a member of the Chinese Academy of Engineering, said the endogenous risk of the system now exceeds that of external attacks. "Now the biggest danger is no longer just hacking; it also encompasses the inherent unexplainability, biases, and hallucinations of algorithms, as well as unauthorized actions triggered by manipulation -- malicious behavior, essentially." He called for shifting from post-incident response to near-instantaneous prediction, with AI detecting highly covert attack precursors amid massive network noise within millisecond-level windows and intercepting advanced persistent threats before they form.

Wang Xin, director of the DAS-Security Research Institute, said the development of AI models has broken the "balance of attack and defense," prompting enterprises to pay more attention to AI security. He added that security business may shift from human-centered detection to AI-centered approaches, with demand growing faster than traditional cybersecurity. According to IDC forecasts, China's AI security revenue will rise from CNY4.41 billion in 2025 to CNY34.03 billion in 2030, a compound annual growth rate of about 50.5 percent, while the broader cybersecurity market is expected to grow at about 8.9 percent annually from 2024 to 2029.