Fabraix

Fabraix develops autonomous red-teaming agents that continuously probe customer-facing AI systems for security, logic and alignment failures. Its main product, Nyx, runs black-box, multi-turn attack campaigns against chatbots, autonomous agents, multi-agent systems, browser agents, voice agents and RL systems without needing access to source code, model weights or internal networks.

Nyx adapts its attacks across turns and input modalities (text, voice, images), runs simulations in parallel, and ships with blueprints tuned to specific domains and system types. It can integrate with CI/CD pipelines for continuous re-testing, and payloads are synthetic with canary secrets to avoid real harm. Fabraix also offers Playground, an open-source adversarial testing environment.

Nyx is aimed at teams building or operating AI agents across industries including finance, healthcare, retail, telecommunications, technology and travel. Enterprise plans include SSO and audit logging, and a Data Processing Agreement is available. Pricing details are available on request.