tools-only A.I.G (AI-Infra-Guard) is a comprehensive, intelligent, and user-friendly AI Red Teaming security testing platform developed by Tencent Zhuque Lab.
Core Features
- AI Infrastructure Vulnerability Scanning: Detects known CVE vulnerabilities in the web services of various frameworks and components (e.g., Ollama, ComfyUI) used in AI training, deployment, and application building, through precise fingerprinting.
- AI Tool Protocol Scan: Utilizes AI Agents to intelligently identify security threats in MCP Servers and Skill, such as tool poisoning and code vulnerabilities, assisting AI tool developers and app marketplaces in completing security certification before release.
- Large Language Model Security Health Check: Employs high-quality and diverse evaluation datasets to automatically evaluates the overall security of different large language models when faced with malicious, adversarial, or deceptive prompt inputs. It exports bad cases to support security alignment and guardrail reinforcement.
Collaboration and Contribution
We invite teams and individual developers to join us in building the leading AI Red Teaming security testing platform in the industry. A.I.G is now open-source on Github and supports rapid configuration of AI component fingerprints, CVE vulnerability fingerprints, jailbreak evaluation sets, and MCP security detection plugins through its plugin management system.
Contact Us
We welcome feedback on platform bugs and contributions to improve detection capabilities. You are invited to join the A.I.G communication group for discussions.