Why is LLM security and prompt injection important in production AI systems?

LLM security is important because models may process untrusted user content, retrieved documents, and tool outputs in the same context.