43. Handle deep learning security risks
Protect models, data, and users from privacy leaks, prompt injection, data poisoning, model theft, adversarial examples, and unsafe outputs. You will learn practical defenses and when security review is needed.