#owasp-llm

Prompt Injection vs. Jailbreaking: Two Conflated Attack Classes

Prompt injection and jailbreaking both use natural language to subvert LLM behavior, but the attacker, the trust boundary that breaks, and the defenses