[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$f4NmRa5gJh0wCnL7fK7eG-_VwlzzVypMZUZKTpVrhK4E":3},{"article":4,"iocs":41,"watch_terms":42,"solutions":43,"lesson":51},{"id":5,"title":6,"slug":7,"summary":8,"ai_summary":9,"brief":10,"url":11,"image_url":12,"published_at":13,"ingested_at":14,"relevance_score":15,"entities":16,"category_id":23,"category":24,"article_tags":28},"f8ef1f61-f04a-4dd6-97ed-fd04e4bb2845","How to Evaluate AI Agent Cybersecurity Capabilities","how-to-evaluate-ai-agent-cybersecurity-capabilities-5a3597","A tutorial for building a benchmark for offensive agents, based on how we built the Wiz Cyber Model Arena","Wiz has launched version 3.9 of its Cyber Model Arena, a benchmark designed to evaluate the cybersecurity capabilities of AI agents. The arena aims to answer whether an agent can perform real-world attack tasks given specific constraints, such as a time limit and no external knowledge sources. The methodology emphasizes building tasks from real-world distributions and grading with code rather than relying solely on LLM judges to ensure realism and prevent cheating.","Wiz launches Cyber Model Arena v3.9 to benchmark AI agent cybersecurity capabilities.","https:\u002F\u002Fwww.wiz.io\u002Fblog\u002Fcyber-model-arena-ai-benchmark","https:\u002F\u002Fwww.datocms-assets.com\u002F75231\u002F1793820803-benchmark-2x.png","2026-11-10T16:03:25+00:00","2026-11-10T17:00:15.323328+00:00",7,[17,20],{"name":18,"type":19},"Wiz","vendor",{"name":21,"type":22},"AI agent","technology","839da5c1-3c34-47e2-9499-f7201640e3ac",{"id":23,"icon":25,"name":26,"slug":27},null,"AI Security","ai-security",[29,34,36],{"category":30},{"id":31,"icon":25,"name":32,"slug":33},"02371804-cf6d-4449-98de-f1a2d4d9b266","Tools","tools",{"category":35},{"id":23,"icon":25,"name":26,"slug":27},{"category":37},{"id":38,"icon":25,"name":39,"slug":40},"e7b231c8-5f79-4465-8d38-1ef13aea5a14","Threat Intelligence","threat-intelligence",[],[18],{"controls":44,"vendors":45},[],[46],{"slug":47,"name":18,"logo_path":48,"roles":49},"wiz","\u002Fvendor-logos\u002Fwiz.png",[50],"reported_by",{"slug":52,"title":53,"body":54,"framework_refs":55,"framework_notes":60},"ai-agent-cybersecurity-benchmarking-highlights-real-world-attack-capability-risks","AI Agent Cybersecurity Benchmarking Highlights Real-World Attack Capability Risks","Wiz's Cyber Model Arena v3.9 reveals that AI agents are increasingly capable of executing real-world cyberattack tasks autonomously, raising the stakes for defenders who must now account for AI-assisted adversaries. The benchmark's methodology—grounding tasks in real-world distributions and using code-based grading—demonstrates that AI attack capabilities are measurable, reproducible, and improving rapidly. Organizations that fail to understand or assess AI-driven threat capabilities risk being caught off-guard by attack patterns that outpace traditional defenses. This matters because the same AI tools being benchmarked for offense can also inform defenders about where their detection and response capabilities are weakest.",[56,57,58,59],"NIST CSF 2.0 ID.RA-01: Asset Vulnerabilities Are Identified and Documented","NIS2 Article 21(2)(f)","CIS Control 18: Penetration Testing (v8)","NIST SP 800-53 CA-8: Penetration Testing",{"NIS2 Article 21(2)(f)":61,"CIS Control 18: Penetration Testing (v8)":62,"NIST SP 800-53 CA-8: Penetration Testing":63,"NIST CSF 2.0 ID.RA-01: Asset Vulnerabilities Are Identified and Documented":64},"Benchmarks like Cyber Model Arena directly address the NIS2 requirement to assess the effectiveness of cybersecurity measures against realistic and evolving threats, including AI-driven ones.","CIS Control 18 calls for realistic adversarial testing, and AI agent benchmarking represents an emerging dimension of that testing that organizations should incorporate into their red-team programs.","This control requires organizations to conduct penetration testing that reflects realistic threat capabilities, which now must include the possibility of AI-automated attack agents.","Understanding AI agents' real-world attack capabilities is a form of risk identification that organizations must incorporate into their vulnerability and threat assessments."]