Explore AI & LLM Security

Large language models and the agentic systems built on them introduce an attack surface most security training never mentions: prompt injection, tool/agent abuse, data exfiltration through context, excessive agency, and model/data poisoning. This topic teaches the OWASP LLM Top 10 and how to threat-model an AI feature the way an attacker would — treating the model as an untrusted interpreter wired to real tools and real data. Explore it with a personal Socratic tutor — you steer, and Nugget follows your thread.