Skip to content
Trending storyDeveloping

GCSA reports 91.3% on CyberGym Level 1 for its security agent

1 article1 sourcesince Nov 7Last article 41m ago ·

Overview

AISummary of 1 article

GCSA says its Agent achieved a 91.3% success rate on CyberGym Level 1, a benchmark in which agents build test programs that reproduce vulnerabilities from descriptions and codebases.

The claim comes from a post by meng shao on X, which reports the score without showing the full evaluation setup.

The same post says GCSA's ecosystem products, covering threat intelligence, asset mapping, AI auditing, edge protection and infrastructure defense, have more than 380 million global users, and that these tools could potentially be called by agents. The post does not show full end-to-end automation of those tools.

Written by AI from the articles below · updated Nov 7, 1:34 AM ET

Check the sources:

Article timeline

The articles in this story. Times are ET.

Nov 7
  1. meng shaoX
    GCSA says its Agent reached 91.3% on CyberGym Level 1 evaluation

    AIGCSA says its Agent achieved a 91.3% success rate on CyberGym Level 1, a benchmark where agents build test programs that reproduce vulnerabilities from descriptions and codebases. GCSA also reports more than 380 million global users for its ecosystem products, which span threat intelligence, asset mapping, AI auditing, edge protection and infrastructure defense. The post describes these tools as potentially callable by agents, though it does not show full end-to-end automation.

    Image from @shao__meng's post

Heat Trend

Not enough continuous observations to show a trend yet.