I built an AI support bot for a fictional food delivery company called Bitely — then broke it five different ways live, including getting it to comp an order it wasn't authorized to refund and leak an internal staff code it was explicitly told to protect. This is prompt injection, the #1 risk on the OWASP LLM Top 10, demonstrated end-to-end with the exact prompts. If you build, secure, or manage anything with an LLM behind it, this is the hour that shows you why "just write a stricter system prompt" doesn't work — and what actually does. BREAK THE BOT — free CTF Watching is one thing. Run these attacks yourself against a deliberately vulnerable bot: https://breakbot.lufsec.com/ GO DEEPER — Introduction to Prompt Hacking for LLMs The live session was the simplest version of every attack. The course covers how to test systematically, chain attacks into a real finding, and write it up so it gets fixed. Two codes: - HACKLLM40 — launch price, first cohort only (expires soon) - YOUTUBE40 — for people who found this on YouTube https://www.lufsec.com/products/courses/prompt-hacking-for-llms-intro Free Prompt Injection Cheat Sheet (every attack from this video): https://www.lufsec.com/products/digital_downloads/llm-security-cheat-sheet ️ WHO THIS IS FOR Security engineers, pentesters, AppSec teams, developers shipping LLM features, and anyone responsible for answering "is our AI safe?" No machine learning background required. Topics: prompt injection, LLM security, jailbreak, indirect prompt injection, AI red teaming, OWASP LLM Top 10, system prompt extraction, AI chatbot security, generative AI risk. LufSec — offensive security training for the AI era. Join the community: https://discord.gg/wHps8Ymm4C ️ For education and authorized testing only. Everything here runs against a bot I built and own. Never test systems you don't have permission to test. #PromptInjection #AISecurity #LLMSecurity #CyberSecurity #AIHacking