Introduction to Prompt Injection Testing

Artificial Intelligence and Large Language Models (LLMs) have rapidly moved into production environments. With this adoption comes a new attack surface: prompt injection.

Prompt injection refers to manipulating the natural language instructions given to an LLM in order to make it behave in unintended ways. This can include bypassing safeguards, exfiltrating sensitive data, or performing unauthorized actions.

In this series, I will share a collection of real-world prompt injection examples. Some are sourced from open communities, others are self-developed during active pentesting work. The goal is to provide defenders and researchers with practical insight into how attackers think, and how security controls can be tested and improved.