As artificial intelligence agents evolve from assistive tools into autonomous decisionmakers and executors, the traditional logic of cybersecurity defense is being fundamentally rewritten, experts and officials said.
Li Bing, vice-chairman of the All-China Federation of Industry and Commerce, said the AI industry is at a critical juncture, transitioning from isolated technological breakthroughs to full-scale value creation.
He called on private enterprises — as key drivers of technological innovation — to actively participate in the “AI plus” initiative, increase research and development investment, and accelerate the large-scale deployment and commercialization of foundation models and agent technologies, while adhering to security guardrails and ensuring that AI benefits humanity. Li made the comments at the 14th Internet Security Conference, also known as ISC. AI 2026, held in Beijing last week.
Ren Xianliang, secretary-general of the World Internet Conference, said: “The development of AI agents must embed values of fairness, inclusiveness and security throughout the R&D and application process, while strengthening international rule-making and multilateral coordination to make AI agents a global public good that safeguards security and benefits all.”
Zhao Zhiguo, vice-chairman of the Internet Society of China and former chief engineer of the Ministry of Industry and Information Technology, called for a holistic approach to national security, urging stronger core technology R&D, improved tiered classification management and assessment mechanisms, and the establishment of security standards covering the entire lifecycle from development and deployment, to operations and maintenance.
Wu Hequan, an academician from the Chinese Academy of Engineering, said that AI agent-driven attacks now exhibit autonomous decision-making, swarm coordination and continuous evolution. He argued that the security paradigm must shift from passive defense to active immunity, building a systematic security operations framework that covers prediction, prevention, detection, response and recovery throughout the entire lifecycle.
Zhou Hongyi, founder of 360 Security Group, warned about AI’s impact on the cybersecurity industry. Using the US AI company Anthropic’s Mythos model as an example, he pointed out that fully automated vulnerability discovery and attack chain construction capabilities have increased vulnerability discovery speed while reducing costs to less than one-thousandth of previous levels.
This, he said, has shattered the 30-year-old offensive-defensive balance built on the assumption that “vulnerabilities are hard to find.”
“In the past, it was about who was stronger; in the future, it will be about who is faster,” Zhou said.
He argued that China cannot simply replicate the foreign approach of relying solely on computing power and model capabilities. Instead, the country should leverage its engineering strengths and pursue an agent-based path — integrating large model capabilities, security expert experience and vulnerability knowledge bases into collaborative agent systems.
The company’s cyber vulnerability mining agent “Tulongfeng” — positioned as China’s answer to Mythos — has already discovered 3,432 vulnerabilities, with 105 confirmed by regulators, covering open-source code, operating systems, office software and AI agent platforms, Zhou said.
On the defensive side, 360 also launched an automated cybersecurity defense system that enables security monitoring, analysis, response and remediation through AI agent collaboration, ushering security operations into the era of “autonomous driving”.
Tanks to chinadaily.com.cn
Please visit:
Our Sponsor