Back to Gap Map Product / UX / Safety & Guardrails

AI-Infra-Guard

Tencent
open source / Overall score: 3.0

AI-Infra-Guard (A.I.G) is a self-hosted red-teaming platform for AI systems from Tencent Zhuque Lab. It scans AI infrastructure for known vulnerabilities, audits MCP servers, agent skills and running agents, and runs jailbreak and safety evaluations against models served through OpenAI-compatible APIs. The platform runs as a web service in Docker, and each scanner is also published as its own Python package.

Openness

5 medium confidence
5.0
license
Apache-2.0 (OSI; the README separately asks integrators to credit Tencent Zhuque Lab)
source
public (the web platform and every standalone scanner build from the repository)
core features withheld
no — no enterprise path in the repository tree and no paid edition named in the README

AI-Infra-Guard's code is Apache-2.0, and the web platform and every scanner build from the public repository. The README also asks anyone building it into another product to credit Tencent Zhuque Lab, which the license itself does not require, and names no paid edition.

Adoption

2 low confidence
2.0

GitHub stars are the only signal used, and a star is not a use. The platform ships mainly as a Docker image whose pull count is a lifetime total, and each scanner package on PyPI covers only one part of it, so neither is counted.

Capability

4 medium confidence
4.0

AI-Infra-Guard covers ground garak does not, scanning the servers, MCP tools and agents around a model as well as the model itself, and its jailbreak evaluation works against any OpenAI-compatible endpoint. garak's library of probes is the deeper of the two on the model alone, so the two sit level.

  • https://raw.githubusercontent.com/Tencent/AI-Infra-Guard/HEAD/README.md recorded 2026-11-12

    The README says A.I.G "brings together OpenClaw security checks, Agent security scanning, AI tool and skill scanning, AI infrastructure vulnerability scanning, and LLM security assessment (jailbreak evaluation)", and documents standalone aig-infra-scan, aig-compliance, aig-model-scan, aig-mcp-scan, aig-skill-scan and aig-agent-scan tools, with fingerprint, vulnerability, MCP and evaluation-set rules under data/.

Verified 2026-11-12