English

← Back to list

Tencent Open-Sources AI Red Teaming Platform, 55 Components

itadol5j

Tencent AI-Infra-Guard

🛡️ Tencent Open-Sources an AI Red Teaming Platform: 55 Components Covered, a Security Testing Powerhouse

Tencent has open-sourced AI-Infra-Guard (A.I.G) — a full-stack AI red teaming platform that covers 55 AI components in one go.

It's the first open-source red team tool to comprehensively cover AI ecosystem security, already used by teams like DeepSeek and showcased at Black Hat EU 25 Arsenal.

🎯 Five Core Capabilities

Feature Description
ClawScan One-click assessment of OpenClaw security risks (config, Skills, CVEs, privacy)
Agent Scan Multi-agent automated scanning framework, supports Dify and Coze
MCP/Skills scanning 14 risk categories, supports source code and remote URLs
AI infra vulnerability scanning 55+ AI frameworks, 1,000+ known CVEs
Jailbreak evaluation Prompt security risk assessment, tested with multiple attack methods

📊 Key Numbers

Metric Value
AI component coverage 55+ AI framework components
CVE vulnerabilities 1000+ Known vulnerabilities
Risk categories 14 major security risk categories
OpenClaw CVE 281 New CVEs/GHSAs

🔍 Covered AI Components

AI-Infra-Guard supports the following mainstream AI frameworks:

Category Components
Inference frameworks Ollama, vLLM, Triton Inference Server, KubeAI
Generative AI ComfyUI, LobeHub, n8n, CrewAI
API services LiteLLM, New-API, Blinko
Agent platforms Dify, Coze, OpenClaw

New in v4.1.3: CrewAI, KubeAI, LobeHub

🚀 Quick Install

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

Visit:http://localhost:8088

Method 2: One-line install script

curl https://raw.githubusercontent.com/Tencent/AI-Infra-Guard/main/docker.sh | bash

Method 3: ClawHub Skill

clawhub install aig-scanner

Configure AIG_BASE_URL to point at your running A.I.G service.

💡 Usage Scenarios

Scenario 1: Scanning a local vLLM instance

  1. Start vLLM (e.g. python -m vllm.entrypoints.api_server)
  2. In the A.I.G web UI, click "AI Infrastructure Security Scan"
  3. Enter http://127.0.0.1:8000
  4. Click "Start Scan"
  5. Review the report: component versions, matched vulnerabilities, severity, fix links

Scenario 2: Scanning an MCP Server

Enter a remote URL (e.g. https://github.com/user/mcp-server) or upload local source code.

Scenario 3: Jailbreak evaluation

In "Settings → Model Config", set the target LLM's API endpoint, pick a dataset, and start the evaluation.

🌟 Highlights

  • Black Hat EU 25 Arsenal: selected for the Black Hat Europe 2025 Arsenal showcase
  • Used by DeepSeek: trusted by the DeepSeek team
  • DeepSeek Awesome Integration: listed in DeepSeek's official integrations
  • ClawHub Skills: EdgeOne ClawScan, EdgeOne Skill Scanner, AIG Scanner
  • Full API: Swagger docs, easy to integrate
  • Multilingual: Chinese and English UI
  • Cross-platform: Linux, macOS, and Windows supported
  • Free and open source: Apache 2.0 license

⚠️ Important Security Notice

AI-Infra-Guard is designed as an AI red teaming platform for internal use by companies or individuals:

  • No authentication: should not be deployed on public networks
  • Internal use only: meant for internal security self-checks
  • Risky operations: scans may trigger security alerts

🔥 Supply Chain Attack Detection

v4.1.1 detected the LiteLLM supply chain attack (CRITICAL severity):

  • Precisely identified the supply chain attack
  • Issued a timely alert for a critical vulnerability
  • Provided remediation guidance

📝 Wrap-Up

Tencent's AI-Infra-Guard is the first full-stack AI red teaming platform from China:

  • 55 AI components covered
  • 1,000+ known CVEs
  • Five core capabilities: ClawScan, Agent Scan, MCP/Skills, AI Infra, Jailbreak
  • Showcased at Black Hat EU 25 Arsenal
  • Used by the DeepSeek team
  • Completely free and open source (Apache 2.0)

For AI security red teaming, Tencent just handed us a seriously sharp tool.

GitHub:https://github.com/Tencent/AI-Infra-Guard

Docs:https://tencent.github.io/AI-Infra-Guard/

Pro version:https://aigsec.ai/(invite code required)