What is prompt injection and how is it mitigated in LLMOps?

Prompt injection is an attack where user input manipulates system instructions; mitigation involves isolation and filtering.

Prompt injection occurs when malicious inputs override or confuse system prompts, causing unintended behavior. In LLMOps, mitigation includes separating system and user prompts, using input sanitization, enforcing structured outputs, and applying security classifiers to detect malicious instructions.