About Us
Focused on AI Security, Safeguarding an Intelligent Future.
Tencent Zhuque Lab is an elite security laboratory established in 2019 under Tencent Security Platform Department, focusing on practical offense-and-defense and cutting-edge research in AI security. Our research covers LLM security, AI agent security, AI-empowered security, and AI-generated content detection. The team has helped renowned vendors such as NVIDIA, Google, and Microsoft, as well as open-source communities including OpenClaw, Linux, and Hugging Face, fix numerous critical vulnerabilities, earning official public acknowledgments.
We have successively launched the open-source AI red-team testing platform A.I.G (AI-Infra-Guard) and the Zhuque AI Detection Assistant, among other AI security products. Our research has been widely published at top-tier international security and AI conferences including Black Hat, DEF CON, ICLR, CVPR, NeurIPS, and ACL, and we have authored the book AI Security: Technology and Practice.
A.I.G (AI-Infra-Guard)
An open-source, comprehensive, intelligent and easy-to-use AI red-team security testing platform by Tencent Zhuque Lab, providing full-stack AI ecosystem security risk self-assessment solutions for enterprise security teams and AI developers. A.I.G has been selected for the Black Hat Arsenal and officially recommended by DeepSeek.
AI Detection Assistant
Zhuque AI Detection Assistant leverages deep learning to accurately identify AI-generated content (text, images, etc.), enabling users to quickly distinguish AIGC content.
AI Sec Matrix
The world's first systematic AI security threat framework, comprehensively mapping security threats and attack paths facing AI systems from an attacker's perspective, providing structured defense guidance.
SecBench
A cybersecurity LLM evaluation benchmark that systematically assesses the comprehensive capabilities of large language models in security knowledge understanding, threat analysis, vulnerability detection, and more.
SkillTrustBench
A premier benchmark co-released by Tencent Zhuque Lab and Prof. Baoyuan Wu's research group at CUHK(SZ) for evaluating AI security scanners' ability to detect malicious agent skills, covering 9 attack categories and 5 agent dependency levels with 5,520 samples.