FreakLabs is a hands-on AI security training platform built by KageX, an offensive AI security startup based in Noida, India. It was created for people who want to learn how AI systems actually get attacked, not just read about it. Instead of slides and multiple-choice quizzes, FreakLabs drops you into live attack labs where you work against real large language models, AI agents and retrieval-augmented generation (RAG) systems running in private sandboxes.
Today the platform includes 20 attack labs and a 9-level curriculum with 54 lessons. The labs cover the attacks that matter most for teams shipping AI products: prompt injection, jailbreaks, data leakage, insecure tool use by agents, poisoned retrieval content and other weaknesses that appear when language models are connected to data, tools and users. Each lab gives you an isolated environment, so you can experiment freely without touching anyone else's systems and without breaking anything that matters.