Artificial Analysis, an independent platform for benchmarking AI models, has entered the cybersecurity sector. On September 25, the company introduced its Cyber Index (v1), designed to measure the effectiveness of AI systems in identifying, validating, and patching security vulnerabilities within enterprise environments.
The index utilizes a composite of three benchmarks developed alongside industry partners: CWE-Bench-AA, DeepsecBench-AA, and CyberGym-E2E-AA. These tools collectively address the entire lifecycle of vulnerability management, ranging from initial discovery to final remediation.
The benchmarks serve distinct functions: CWE-Bench-AA utilizes Common Weakness Enumeration categories, DeepsecBench-AA performs advanced security analysis, and CyberGym-E2E-AA tests AI agents through end-to-end scenarios on realistic attack surfaces.
A key feature of the index is its focus on defensive cybersecurity. The platform evaluates how AI models assist businesses in preemptive vulnerability management rather than assessing their potential to compromise systems.
Artificial Analysis, co-founded by Micah Hill-Smith, previously released indices for the finance, legal, and healthcare sectors. Additionally, the company maintains a general-purpose Intelligence Index, which is currently at version 4.3.2 and covers 10 evaluations across hundreds of models.
