5 min read
Defensive Techniques A practical framework for implementing prompt injection detection and containment at the API gateway layer — covering input sanitisation, context isolation, output filtering, and anomaly-based detection for production LLM deployments.