A multinational organization has set up automated AI-driven pipelines to support its customer service operations.

According to the EC-Council CAIPM framework, the AI infrastructure stack is typically divided into multiple layers, including the foundation layer, compute layer, data layer, and AI/ML platform layer. Each layer has distinct responsibilities, and identifying issues correctly depends on understanding what each layer governs.

In this scenario, the problems are related to authentication rules, network routing, and security controls. These are not related to data quality, model logic, or AI tooling. Instead, they are core infrastructure components that define how systems communicate, how access is controlled, and how environments are secured. These elements fall squarely within the foundation layer, which includes networking, identity and access management, security policies, and environment consistency across development, testing, and production.

The key clue in the question is that the AI models and tools remain unchanged, yet failures occur only in production environments. This indicates that the issue is not in the AI/ML platform or compute execution but in the underlying infrastructure that supports deployment and runtime operations. CAIPM emphasizes that scalable AI systems require stable, standardized foundational infrastructure before higher-level AI capabilities can function reliably.

Therefore, since the inconsistencies arise from differences in networking, authentication, and security configurations across environments, the correct answer is Foundation layer, as it directly governs these foundational infrastructure elements.