Fabraix is an AI security research and red-teaming platform that tests AI agents for security, logic, and alignment weaknesses.

Nyx is Fabraix's autonomous adversarial testing agent for probing AI systems through multi-turn black-box interactions.

Does Fabraix require source code?

No. Nyx is designed for black-box testing and can assess an exposed AI system without source code or model weights.

Nyx can test text, voice, browser, document, research, and multi-agent interaction surfaces for adversarial failure modes.

How is Fabraix priced?

The current platform offers free research access by application, while production scans and continuous testing use custom or usage-based pricing.

Fabraix

Autonomous AI red-teaming platform to detect safety vulnerabilities.

Visit Website

What is Fabraix?

TL;DR

Fabraix is an autonomous AI red-teaming platform that uses adaptive adversarial testing to uncover security, logic, and alignment weaknesses in customer-facing AI agents before those failures become production risks.

Flagship red-teaming agent Nyx launching adaptive adversarial attacks
Black-box penetration testing requiring only target endpoint URLs
Continuous CI/CD integration automatically re-testing model updates

Explore with AI

About

Fabraix is an AI security research and testing platform focused on adversarial verification for AI systems. Its Nyx product probes AI agents and applications through black-box, multi-turn testing to uncover weaknesses before attackers or users encounter them. The platform is built for AI developers, security engineers, internal red teams, and organizations deploying customer-facing agents. Nyx can test systems without requiring source code, model weights, or deep internal integration, making it useful for evaluating exposed agent surfaces. Fabraix emphasizes adaptive testing across multiple interaction surfaces including text, voice, browser, and document workflows. Its testing can run many simulations in parallel and generate findings that security teams can reproduce and use for remediation. For FutureStack, Fabraix belongs in Developer Tools because it provides infrastructure for AI security verification. It is best for technical teams that need repeatable red-team coverage for AI agents and model-powered applications. Production pricing is based on the scope of scans or continuous testing.

Use Cases

Red-Team AI Agents

Run adaptive adversarial scenarios against customer-facing agents to discover jailbreaks, prompt injection, unsafe behavior, and logic failures.

Test AI Releases

Repeat security evaluations when models, prompts, tools, or agent behavior changes so new vulnerabilities are caught before deployment.

Validate Guardrails

Probe AI systems with challenging inputs and multi-turn attack paths to evaluate whether safety controls hold under pressure.

Support AI Security

Generate reproducible findings and remediation guidance that security and engineering teams can use during AI risk reviews.

Key Features

Flagship red-teaming agent Nyx launching adaptive adversarial attacks
Black-box penetration testing requiring only target endpoint URLs
Continuous CI/CD integration automatically re-testing model updates
Self-play adversarial loops generating and retaining new jailbreaks

Featured Tools

Wispr Flow

Voice-First Dictation

We may earn commissions from links to support our work. Learn more.