A comprehensive AI red teaming platform developed by Tencent Zhuque Lab, featuring AI infrastructure vulnerability scanning, MCP server risk assessment, and jailbreak evaluation capabilities to help organizations conduct self-assessment of AI security risks。
One-Minute Overview#
A.I.G is an AI red teaming platform developed by Tencent Zhuque Lab, designed for organizations and individuals to comprehensively detect AI system security risks. If you need to evaluate AI infrastructure, MCP servers, or prompt security, A.I.G provides a professional and user-friendly solution.
Core Value: A one-stop AI security testing platform covering comprehensive security assessment from infrastructure to model prompts
Quick Start#
Installation Difficulty: Low - Docker-based containerized deployment with three simple installation methods
# Method 1: Docker Compose (Recommended)
git clone https://github.com/Tencent/AI-Infra-Guard.git
cd AI-Infra-Guard
docker-compose -f docker-compose.images.yml up -d
Is this suitable for me?
- ✅ Organizational internal security assessment: Ideal for enterprises conducting comprehensive security testing of self-built AI systems
- ✅ AI model developers: Quickly evaluate model security and jailbreak risks
- ✅ MCP server maintainers: Detect potential security vulnerabilities in servers
- ❌ Public network deployment: Currently lacks authentication mechanism and should not be deployed on public networks
Core Capabilities#
1. AI Infrastructure Scan - Comprehensive Vulnerability Detection#
- Precisely identifies 30+ AI framework components and covers 400+ known CVE vulnerabilities, including Ollama, ComfyUI, vLLM, etc. User Value: Prevents AI system security incidents caused by underlying component vulnerabilities
2. MCP Server Scan - Intelligent Risk Detection#
- Powered by AI Agent technology, detects 9 major categories of MCP security risks, supports source code/remote URL scanning User Value: Early detection of MCP server security threats to prevent data breaches and unauthorized access
3. Jailbreak Evaluation - Prompt Security Assessment#
- Rapidly evaluates prompt security risks, includes multiple curated jailbreak evaluation datasets, supports cross-model security performance comparison User Value: Prevents models from being manipulated by malicious prompts, ensuring AI system security and reliability
Technology Stack & Integration#
Development Languages: Python, JavaScript, HTML, CSS, Shell, Dockerfile Main Dependencies: Docker Compose, Swagger, React (UI) Integration Method: Web Interface / API
Maintenance Status#
- Development Activity: Actively developed with recent v3.5-preview-2 release adding 100+ AI component CVEs
- Recent Updates: Regular updates with continuous addition of new vulnerability databases and evaluation features
- Community Response: Active GitHub discussions and issue response mechanism encouraging community contributions
Documentation & Learning Resources#
- Documentation Quality: Comprehensive
- Official Documentation: https://tencent.github.io/AI-Infra-Guard/
- Example Code: Provides API documentation and complete example code
- Getting Started Guide: Detailed onboarding process with Chinese and English documentation support