Large Language Models
StruQ and SecAlign Crush Prompt Injection Attacks to Near-Zero
Prompt injection just got hammered. Two new defenses, StruQ and SecAlign, slash attack success to near zero across major LLMs, preserving performance.