TrendAI™ has claimed the top position on CyberGym, an independent University of California, Berkeley benchmark for AI-driven security tools. Its agentic exploit-remediation engine, codename AESIR, achieved a 97% success rate, surpassing all other entrants and outperforming the previous leader’s 93.2% score.
CyberGym evaluates AI security solutions against 1,507 confirmed vulnerabilities from 188 major open-source projects. TrendAI™ secured the lead by combining seven AI models from Anthropic, DeepSeek, Google and OpenAI, with Claude Opus 4.6 serving as the primary engine. The system draws on more than 20 years of vulnerability research and exploit intelligence from TrendAI™ ZDI to identify, validate and remediate threats faster than attackers can exploit them.
“Cybersecurity has long been defined by the scramble from discovery to exploitation,” said Sharda Tickoo, Country Manager for India and SAARC at TrendAI™. “Our job is to make that gap meaningless by compressing exposure time towards zero.”
Key highlights of the benchmark results include:
TrendAI™ ranks first, more than 12 points ahead of GPT-5.6 Sol and Claude Mythos 5
The platform uses a persistent vulnerability ontology with over 12,500 episodic memories and 15,500 exploit seeds
Nearly one-third of exploits were resolved without AI involvement, while classical fuzzing outperformed AI reasoning on roughly a quarter of tasks
TrendAI™ Vision One™, the company’s unified cybersecurity platform, powers this capability by centralising risk exposure management across AI, cloud, networks and endpoints.