Preparing Your Data Lakehouse for the EU AI Act: Auditable Lineage and Data Provenance

This article provides technical guidance on how to prepare a data lakehouse architecture for the EU AI Act, emphasizing auditable lineage and data provenance. It argues that compliance should be an engineering capability, not just a paperwork exercise. The piece covers how open table formats, catalog metadata, semantic layers, query logs, access controls, and lineage systems can turn regulatory requirements into platform features. It addresses the challenges of fragmented AI data paths, the need for human oversight hooks, accuracy through data contracts, and security for AI agents. A practical readiness checklist and the importance of explainable operations are also included, with a focus on agentic analytics scenarios.