OWASP LLM
LLM05 · Improper Output Handling
Model output is passed to browsers, shells, SQL or other systems without validation (XSS, SSRF, RCE).
OWASP Top 10 for LLM Applications (2025)
Mitigations
- Treat model output as untrusted user input
- Context-aware encoding and parameterised queries
- Sandbox code execution
Detections 1
Related techniques
- ASI05 Unexpected Code Execution OWASP Agentic
In the news
No stories mapped to this technique yet.
Tools
Test or exercise it
- ai-exploits Model security · Hold · Offense
Detect or mitigate it
- Guardrails AI Guardrail · Trial
- LLM Guard Guardrail · Hold
Mappings are AI-assisted and checked against the framework definitions; corrections welcome.