Adversarial AI red-teaming agents that continuously detect security vulnerabilities in customer-facing AI systems.
Fabraix builds AI red-teaming agents that autonomously probe customer-facing AI agents to find security vulnerabilities. Its flagship agent Nyx achieved a 78% attack success rate on the AgentHarm benchmark compared to 67% for GPT-5.6 Sol, and has found vulnerabilities at dozens of Fortune 500 companies. The platform is designed for continuous testing because AI agents can change behavior with every new model, prompt, tool, permission, or data source update. Fabraix is a product of Fabraix.