CrowdStrike and Meta launch CyberSOCEval, an open source benchmark for testing AI in SOCs, setting new standards for secure adoption of LLMs in cybersecurity.
CrowdStrike and Meta have unveiled CyberSOCEval, an open source benchmark suite designed to evaluate how large language models (LLMs) perform in real-world security operations.
Built on Meta’s CyberSecEval and enhanced with CrowdStrike’s threat intelligence, CyberSOCEval measures AI performance across incident response, malware analysis, and adversary simulations. The framework aims to give organisations clearer metrics for adopting AI in security operations centres (SOCs) while providing developers with a standardised benchmark to improve model capabilities.
“Our collaboration with CrowdStrike introduces a new open source benchmark suite to evaluate the capabilities of LLMs in real-world security scenarios,” said Vincent Gonguet, director of product, GenAI at Superintelligence Labs at Meta.
At its Fal.Con event in Las Vegas, CrowdStrike also announced broad integrations with Amazon Web Services, Intel, Meta, NVIDIA, and Salesforce, extending its Falcon platform across the enterprise AI landscape. These collaborations bring Falcon security into AWS SageMaker, Bedrock, and Marketplace; Intel’s NPU-powered AI PCs; Meta’s evaluation tools; NVIDIA’s LLM workflows; and Salesforce’s Agentforce platform.
According to Daniel Bernard, chief business officer at CrowdStrike, “Securing AI is not just about technology – it’s about securing the full ecosystem where AI is built, deployed, and used.”
The company further highlighted findings from Canalys, showing partners can generate up to US$7 in services revenue for every US$1 spent on the Falcon platform. Falcon Next-Gen SIEM emerged as a major driver of recurring services, with nearly half of the services value attributed to managed offerings.
“Canalys’ findings reinforce how the Falcon platform is in a league of its own, setting a new standard for partner opportunity,” Bernard added.
