An AI security platform that detects and blocks prompt injection, jailbreak attempts and sensitive-data leakage in LLM applications, sitting between an app and the model to screen both directions of traffic.