Context information
OWASP LLM01:2025: prompt injection across trust boundaries
OWASP distinguishes direct prompt injection from instructions embedded in external material that a model consumes. The guidance discusses separating trusted instructions from untrusted content, limiting privileges and evaluating adversarial inputs. Retrieved documents and tool responses require their own trust boundary.
- Provenance
- OWASP Gen AI Security Project guidance
- Updated
- 2026-11-08T22:15:33.333306Z