Context information

OWASP LLM05:2025: improper output handling

Model output can become dangerous when downstream components interpret it as executable code, database queries or browser markup. OWASP describes validation, context-aware encoding and parameterized database operations as mitigations. The receiving system remains responsible for handling generated output according to its destination.

External source ↗

Provenance: OWASP Gen AI Security Project guidance
Updated: 2026-11-08T22:15:33.674698Z