ConceptAI & ML
Prompt injection
A vulnerability where an attacker manipulates inputs to make a model execute unintended instructions.
Know first
Named as a prerequisite by the lectures that define this. The number is how many say so. A grey one has no page yet.
Opens up
Concepts that name this one as a prerequisite.
Lecturers
2Each in their own terms, from the easiest level.
Andrej Karpathy
An attack where hidden instructions within user-provided data hijack the LLM's intended prompt behavior.Jim Harris
A technique where carefully crafted user inputs shift token probabilities to override model safety safeguards, privacy rules, or execution restrictions.
Lectures that assume it
1Connections
6The lectures that define it or assume it, and the concepts next to it.