Beyond Guardrails: How to Secure Your AI from Unsanitized Data and Deceptive Prompts