The OWASP Top 10 for LLMs lists the ten most critical security risks for AI-powered applications. This guide walks through prompt injection, insecure output handling, sensitive data leaks, and excessive agency — with practical Python code you can apply immediately to harden your LLM-powered app.