AI OSINT
A curated collection of Google dorks, Shodan and Censys queries, GitHub dorks and Sigma rules for finding exposed AI infrastructure on the internet, aimed at red-team operators, penetration testers and OSINT researchers.
AI OSINT collects reconnaissance techniques for locating exposed LLM endpoints such as Ollama and vLLM, leaked AI API keys, misconfigured vector databases such as Qdrant and ChromaDB, unprotected AI agent gateways such as MCP servers, exposed MLOps platforms such as MLflow and Jupyter, and indexed AI chatbot conversation leaks.
It also documents MCP supply-chain attack patterns, including marketplace poisoning and STDIO remote code execution, AI IDE exploitation such as unsafe autonomous modes and zero-click issues, and known AI-related data-exposure incidents, giving red-team operators and OSINT professionals specific queries and tools for this kind of reconnaissance.
It is presented as a curated list rather than runnable software.