AI & LLM Security
Break the guardrails: prompt injection, system-prompt leaks and jailbreaks — on a bot you build yourself.
One-time payment · lifetime access · lessons, labs, exam and certificate included · no subscription.
About this course
Large language models cannot tell trusted instructions from attacker text in the same stream, which makes prompt injection the top entry on the OWASP LLM Top 10. In this course you build your own guarded chatbot, plant a fake secret in its system prompt, and then learn to break it. You run direct prompt injections that override the rules, coax the model into leaking its own system prompt, and use persona jailbreaks like developer mode. A key lesson covers indirect injection, where malicious instructions hide in content the model reads, such as a document or email an agent processes. Each attack is matched with real mitigations: keeping secrets out of the context, least privilege for tools, treating model output as untrusted, and human approval for high-impact actions. You only ever test the bot you built.