AI Security
What is Prompt Injection?
Prompt injection is an attack against applications built on large language models in which crafted input manipulates the model into ignoring its intended instructions or performing unintended actions. It matters because it can lead to data leakage, unauthorized actions, or bypassed safety controls in AI-powered features.
Example
A document processed by an AI assistant contains hidden text instructing the model to reveal its system prompt, and the model complies.
Reference: OWASP ↗
Related Clear Infosec services
Related terms
Have a question about prompt injection in your environment? Talk to our team.