AI

This course takes the application-security lens to agentic AI systems, assuming you already know how agents, tool use, and function calling work. It goes deep on prompt injection: what it structurally is, why it resists the kind of clean fix that solved SQL injection decades ago, how attackers exploit tool-connected agents to cause real-world harm, and the layered defenses (privilege separation, untrusted-content handling, human-in-the-loop checkpoints, monitoring) that practitioners use to manage a risk that currently has no complete solution.