OWASP LLM

Prompt Injection Is a Boundary Problem, Not a Prompt Problem

Why prompt defenses need independent authority, data-flow and effect boundaries, with explicit enforcement assumptions and evidence limits.