CYBERSECURITYTRACKER
TRACKING6,506 stories in this site build1,309 vulnerability news stories in this site build
Permanent story citation

Dream’s Hero scores 96.6% on CyberGym to rank first globally in autonomous cybersecurity research benchmark

This page keeps the story as Cybersecurity Tracker first published it. If the tracker later corrects it, the correction appears below the original and never replaces it.

Back to newsStory 6829

As cited

Copy frozen at (site build).

ai security

Dream’s Hero scores 96.6% on CyberGym to rank first globally in autonomous cybersecurity research benchmark

Dream announced that its autonomous cybersecurity research system, Hero, scored 96.6% on CyberGym, a UC Berkeley benchmark that evaluates artificial intelligence (AI) systems against over 1,500 real-world security challenges. Hero ranks first globally on the benchmark, outperforming previously published results from leading frontier and cybersecurity-specialized AI systems. The system combines a specialized cybersecurity model called Hercules with orchestrated AI agents and security tools to perform practical vulnerability research tasks including code analysis, exploit development, and remediation.

Why it matters: Government and critical infrastructure security teams should evaluate whether specialized AI-driven vulnerability research tools can augment or replace manual analysis workflows, particularly for organizations operating under sovereign or air-gapped constraints.

Source published
First seen by Cybersecurity Tracker

Source attribution

Correction

Correction recorded as of .

ai security

Dream’s Hero scores 96.6% on CyberGym to rank first globally in autonomous cybersecurity research benchmark

Dream announced that its Hero autonomous cybersecurity research system achieved 96.6% on CyberGym, a UC Berkeley benchmark testing practical vulnerability research across 1,500 real-world security challenges. Hero orchestrates multiple specialized artificial intelligence (AI) agents and security tools, including Hercules, Dream's custom cybersecurity model trained on real vulnerability research data. The system operates entirely within controlled environments without relying on external application programming interface (API) calls or internet access.

Why it matters: Government and critical infrastructure teams should evaluate whether autonomous AI-driven vulnerability research tools can enhance their cyber defense capabilities while maintaining sovereign control over security operations and sensitive data.

Source published
First seen by Cybersecurity Tracker

Source attribution

Glossary