Home

AIX — AI Red Team Framework

AIX is an automated security testing framework for AI/LLM endpoints. It gives penetration testers and red teamers the tools to assess AI systems across the full attack surface — from reconnaissance to exploitation to reporting.

---

Pages

• Installation — pip, source, dev setup
• Modules — full reference for every attack module
• Payload Schema — JSON format, fields, level/risk system
• Adding Modules — how to build a new attack module
• Attack Chains — YAML playbooks, syntax, built-in playbooks
• Database & Reporting — sessions, conversations, HTML export

---

Quick Start

pip install aix-framework

# Recon first — detects model, guardrails, structure
aix recon https://api.target.com/chat -k sk-xxx

# Attack — bypass engine activates automatically if guardrail was found
aix inject https://api.target.com/chat -k sk-xxx

# Full scan
aix scan https://api.target.com/chat -k sk-xxx

---

Architecture

CLI (aix/cli.py)
  → Module (aix/modules/<name>.py)
    → BaseScanner (aix/core/scanner.py)
      → Connector          — sends HTTP / WebSocket / raw requests
      → AIEngine           — LLM-as-a-Judge + payload generation
      → BypassEngine       — session-aware guardrail evasion
      → AIXDatabase        — SQLite persistence
      → Reporter / Finding — structured output + OWASP/ATLAS tags

Every scan is grouped into a session by target. Guardrail data detected by aix recon is stored in the session and automatically consumed by subsequent attack modules.