Indirect Prompt Injection via Retrieved Documents
Retrieved documents have become an undefended attack vector for LLMs.
Marco Quist
Section
Prompt Injection
7 stories in Prompt Injection.
Input Sanitization Pipelines for LLM Security
Prompt injection is a structural risk that requires layered defenses, not a one-time fix.
Selah Beck
Guard LLM Approaches to Prompt Injection Detection
Guard LLMs catch injections but fail fast when attackers add Unicode or tweak inputs slightly.
Bianca Espinoza
Prompt Inversion and Prompt Extraction Risks
Distinguishing two distinct attacks that steal or reverse-engineer a model's hidden instructions.
Selah Beck
Prompt Injection Attacks in Production Systems
Attackers exploit LLM blindspots to breach enterprise systems.
Hugo Dvorak
Prompt Injection vs Jailbreak
Prompt injection and jailbreaking are distinct attacks requiring different defenses.
Bianca Espinoza
Prompt Injection Detection Methods
Effective defenses stack multiple layers since no single method catches all injection attacks.
Selah Beck