At the 2026 Global Forum on Public Safety Cooperation (Lianyungang) held recently, Zhou Hongyi, founder of 360 Group, delivered a keynote speech highlighting that artificial intelligence has fundamentally transformed the dynamics of cyber offense and defense. He introduced a novel automated attack-defense framework built around the principle of "model-countering-model" and disclosed the latest operational metrics for 360's two flagship products, codenamed "Yitian" and "Tulong."

Zhou pointed out that AI is delivering a "three-tier disruptive impact" on traditional cybersecurity. First, AI can autonomously discover vulnerabilities with capabilities surpassing human experts. Second, AI leverages these vulnerabilities to execute automated network attacks. Third, AI can bypass human-defined rules, permissions, and boundaries, introducing risks of losing control. He emphasized that vulnerabilities serve as the critical linchpin and most vital strategic asset in cyber conflict. In the past, vulnerabilities were scarce, allowing a fragile equilibrium between attackers and defenders. Now, AI transforms vulnerability discovery from a "craftsman's skill" into a 24/7 industrialized operation, accelerating discovery speeds by a hundredfold and slashing costs to one-thousandth of previous levels. This heralds an unprecedented global surge in vulnerabilities.

Furthermore, hackers' expertise and capabilities are being "distilled" into large-model skill packs, enabling the creation of hacker agents that can be replicated at scale and operate around the clock. Once these agents infiltrate enterprises and gain access to tools and permissions, they create new internal attack surfaces, along with risks of error propagation between agents and unauthorized collaborative boundary-crossing. To address these evolving challenges in the AI era, 360 has proposed a "model-countering-model" approach, building an AI security framework that encompasses vulnerability discovery, automated defense, and agent governance. Specifically, the "China Version of Mythos," named Tulongfeng, has uncovered over 10,000 vulnerabilities since its June launch, including long-dormant zero-day flaws in Windows and Office, as well as high-value vulnerabilities in AI platforms like OpenClaw and Codex. Meanwhile, Yitiazhen employs a swarm of security agents to autonomously execute threat detection, analysis and verification, coordinated response, and remediation on a 24/7 basis.

These achievements are underpinned by two core capabilities: 360's self-developed foundational security large model and collaborative multi-agent swarm technology. In the AI era, cybersecurity risks span digital infrastructure, large models, and intelligent agents, making it impossible for any single enterprise or industry to tackle them alone. Zhou announced that 360 has launched the "Panshi Shield" security collaboration initiative, uniting enterprises across the IT innovation infrastructure sector to jointly identify vulnerabilities, strengthen defenses, and manage AI-related risks. He expressed willingness to work with all partners to fortify the public security foundation in the age of AI.