How Did China Cellular’s AI Agent Rating on a Brutal World Cybersecurity Benchmark?
A synthetic intelligence cybersecurity agent collectively developed by China Cellular has achieved a 96.35% vulnerability copy success charge on a outstanding worldwide benchmark, rating first globally amongst telecommunications operators and state-owned enterprises, in line with state media stories.
The offensive and defensive AI system, named “TieYing,” or Iron Eagle, was developed by China Cellular Group’s Community Division in collaboration with the Community Administration Middle of its provincial subsidiary, Zhejiang Cellular, the Qianjiang Night Information boasted Wednesday.
The outcome locations China Cellular’s AI-native safety expertise amongst prime international performers at a time when business analysts say digital transformation and cloud adoption are quickly increasing cyberattack surfaces. As automated assaults and vulnerability iterations speed up, cybersecurity specialists word that digital protection is shifting from human-versus-human confrontations to automated battles between opposing AI methods.
TieYing’s efficiency was recorded on the most recent leaderboard for CyberGym, an authoritative AI offensive and defensive analysis framework created by researchers on the College of California, Berkeley, and acknowledged by main worldwide tutorial conferences.
Designed to check the real-world safety capabilities of autonomous AI brokers, CyberGym attracts on a big database of historic vulnerabilities from open-source software program initiatives to simulate genuine cyber battle situations. The benchmark evaluates members on code comprehension, logical reasoning and vulnerability evaluation, attracting entries from main international expertise companies together with Microsoft, Google, OpenAI and Alibaba.
Builders stated TieYing’s rating stems from a specialised multi-agent collaborative structure constructed to beat conventional bottlenecks in vulnerability evaluation, comparable to prolonged investigation cycles, advanced code logic and weak suggestions loops.
On the core of the platform is a worldwide scheduling system pushed by a Meta-Agent, which breaks down advanced safety duties into granular subtasks and executes them in parallel throughout a number of threads to hurry up evaluation.
To stop blind spots and false positives frequent in single-path reasoning, the system cross-validates findings throughout 5 specialised analytical modules: taint monitoring, signature recognition, enterprise logic inspection, provide chain danger analysis and vulnerability sample looking.
Supported by a hierarchical vulnerability data base, dynamic atomic software invocation and full-process state reminiscence, the agent can maintain iterative reasoning over prolonged process cycles, retaining legitimate investigative leads whereas discarding lifeless ends to autonomously uncover and reproduce software program flaws.
Not like general-purpose massive language fashions that target scaling parameter counts, China Cellular engineered TieYing particularly for sensible safety manufacturing, emphasizing correct code studying, exact flaw localization and dependable exploit verification.
China Cellular plans to transition TieYing from benchmark testing into dwell community safety operations. Upcoming deployments will concentrate on automated code auditing, pre-launch danger inspections, legacy vulnerability opinions and routine menace assessments to realize early detection and closed-loop remediation, decreasing reliance on handbook safety operations.

