804Course · Practitioner
Jailbreaks
Bypassing the guardrails.
Bypassing the guardrails. The taxonomy, the automated attacks (PAIR, GCG), and the defenses that actually help.
- 01
Categorize jailbreaks.
- 02
Read the leading papers.
- 03
Design layered defenses.
- 04
Run a jailbreak audit.
Prerequisites
- — Prompt Injection.
Pacing
~3 hours.
- IModule
The category
Precisely.
A Note from the Faculty
Jailbreaks are cheap to find and expensive to prevent.