Chinese AI Solution DoGNAVY Ranks Third in International Security Benchmark

The Chinese AI solution DoGNAVY achieved third place in the international CyberGym AI security benchmark, released on August 5. Developed jointly by China's leading AI research institutions and security teams, the solution also ranked first among open-source offerings with a 90.8% pass rate.

CyberGym is recognized as one of the largest benchmarks for AI agent cybersecurity testing. DoGNAVY's success, co-developed by DARKNAVY, is notable for its high ranking achieved using a single open-source general model (GLM-5.2). This contrasts with higher-ranked competitors like Microsoft and Google, who reportedly utilize multiple closed-source models, routing tasks to the best-performing ones.

DoGNAVY's approach demonstrates that effective AI security solutions can be built using open-source components. This represents a significant development, potentially making advanced security capabilities accessible to a wider range of innovators.

The report suggests that AI security offense and defense are entering a stage of practical application. The open-source route may accelerate this progress and foster improved security within AI systems.