Course
AI Safety & Security
Defending AI systems, and making capable models behave as intended.
L3 · AdvancedFast-movingKnown~3 h
What you’ll learn
- Identify the main attack surfaces of an AI system and their mitigations
- Design layered guardrails and sandboxing around a model or agent
- Explain how models are aligned and why oversight remains necessary
Prerequisites
applied-llm-systems
Module 1. Attack Surfaces
Injection, leakage and poisoning, and excessive agency.
Module 2. Defenses & Alignment
Guardrails, sandboxing, alignment, and oversight.