Prompt Injection vs Jailbreak: A Defense Checklist for Local RAG and Agents
Prompt injection vs jailbreak: a table of attacker, target and defense focus, the lethal-trifecta risk check and a layered defense checklist for local RAG.
-
-
LLM Jailbreak Explained: Jailbreaking, “Po Jia” and “Cracking” Compared, With Principles and Risks
What LLM jailbreak means, how it differs from uncensored models and prompt injection, how it works, how it is defended, and the legal risks. No working prompts.
-
A Short History of LLM Jailbreaks: From 2022 Prompt Injection to 2025 Constitutional Classifiers, and How Attack and Defense Evolved
A short history of LLM jailbreaks: prompt injection, role-play, automated search, long-context and multi-turn attacks, then classifiers. Concepts only.
-
Jailbreak Defenses Explained: Alignment Training, Classifiers and Permission Isolation as Layers of Protection
Jailbreak defenses explained: alignment training, circuit breakers, classifiers, streaming moderation and permission isolation, with their costs and odds math.
