Check Point Lakera Secure LLM Reference Architecture

The Secure LLM Reference Architecture engagement from ControlPlane enables organisations to design secure, scalable, and operationally sustainable AI application architectures on AWS. The service is designed for enterprises adopting Large Language Models (LLMs), Retrieval-Augmented Generation (RAG), copilots, and agent-based AI systems that require clear security boundaries, runtime protections, and integration with existing cloud-native controls.

Through a structured architecture-led engagement, ControlPlane reviews current-state AI platforms and defines a target-state architecture that incorporates runtime security controls, identity and access management, data protection, logging, monitoring, secrets management, policy enforcement, and incident response considerations. The engagement also defines how Lakera integrates into the customer environment to provide protections such as prompt injection prevention, sensitive data protection, and runtime policy enforcement.

The engagement is suitable for AWS-native, Kubernetes, hybrid-cloud, and regulated enterprise environments. Deliverables include reference architecture diagrams, trust boundary mapping, Lakera integration patterns, deployment guidance, and implementation backlogs to support pilot and production rollout activities.

This offering is designed to support AWS services and architectures including Amazon EKS, Amazon Bedrock, AWS IAM, Amazon CloudWatch, AWS Lambda, Amazon API Gateway, AWS Secrets Manager, Amazon GuardDuty, and other cloud-native AWS security and observability services.

Relevant keywords: Secure LLM Architecture, AI Security, Runtime Protection, Lakera, Prompt Injection Protection, RAG Security, Agentic AI Security, Kubernetes Security, Cloud-Native Security, AWS AI Security.