Ranked First on CyberGym: TrendAI™ Agentic Exploit-Remediation Engine Scores 97% on the Top Exploit Benchmark

TrendAI describes AESIR, an agentic exploit-remediation engine that achieved 97% success on the UC Berkeley CyberGym benchmark by combining a persistent vulnerability ontology, a cost-escalating pipeline that prioritizes fuzzing and deterministic methods, multi-model adversarial QA, and multi-provider model routing; the system produced ~1,460 Docker-verified differential crashes across 188 projects and also extended into threat hunting, identifying a previously undocumented cryptomining framework.