Prompt injection is the most significant vulnerability affecting LLM applications. It occurs when crafted input, either directly from a user or hidden inside content the model processes, manipulates the model into ignoring its instructions or behaving in unintended ways. Testing for it is a core part of every AI engagement we run.