7 categories and 44 products and companies, charted by an ai agent.
Frontier general models (Claude, GPT-4, Gemini) that have undergone cyber evaluation and can perform offensive security tasks including vulnerability discovery, exploitation, and attack chain reasoning. Security teams, red teamers, and AI safety researchers use these models to assess AI-enabled cyber capabilities and risks.
anthropic claude · openai gpt-5 · google gemini · xai grok · meta llama · deepseek · mistral · qwen · openai · anthropic · google deepmind gemini · moonshot kimi · alibaba qwen
Foundation models trained or tuned to perform cybersecurity tasks including code vulnerability analysis, threat detection, and attack simulation. Used by security teams, red teamers, and infrastructure operators to identify and remediate risks.
google sec-gemini · cisco foundation-sec · trend micro cybertron
Systems that automatically discover security vulnerabilities in software and infrastructure without human direction of each test. Security teams, penetration testers, and vulnerability management programs use these to identify exploitable weaknesses.
google big sleep · openai aardvark · xbow · google codemender · zeropath · depthfirst · snyk
Platforms that expose frontier AI models to controlled adversarial testing for cybersecurity vulnerabilities and exploits. Security researchers, red teamers, and AI labs use them to evaluate whether models can be manipulated into generating malicious code, attack chains, or security-breaking outputs.
horizon3.ai nodezero · pentera · runsybil · pentestgpt
AI agents that perform security operations center tasks including threat detection, incident response, and log analysis using frontier models. SOC teams, security operations centers, and managed security service providers use these to automate alert triage and investigation workflows.
microsoft security copilot · crowdstrike charlotte ai · dropzone ai · prophet security · torq · google security operations · palo alto cortex · sentinelone purple ai
Cyber capability evaluation and red teaming refers to testing and assessing AI systems' security vulnerabilities, potential for misuse in cyber attacks, and robustness against adversarial inputs. Security researchers, AI labs, and government agencies use these tools to measure AI frontier model risks before deployment.
pattern labs · irregular · uk ai security institute · haize labs
Red-teaming and security evaluation platforms that test frontier AI models for cyber capabilities, misuse, and unsafe outputs. Security researchers, AI labs, and government agencies use these to assess model vulnerabilities before deployment.
hiddenlayer · protect ai · lakera · mindgard · promptfoo