Secures large language model applications against prompt injections and other AI-specific vulnerabilities.