Back to Glossary
What is Prompt Injection?
Definition
An attack against LLM-powered applications where malicious user input manipulates the model's system prompt to execute unintended instructions.
Security Impact
Can lead to data exfiltration, bypassing safety guardrails, unauthorized actions in agentic systems, and complete compromise of AI-powered workflows.
How to Remediate
Implement strict input sanitization, use system prompt isolation, apply output filtering, and never trust LLM outputs as executable commands without validation.
Want to automatically detect and block Prompt Injection vulnerabilities in your CI/CD pipeline?
Start scanning with Bryxe