Element-level metadata is classification and governance metadata attached below the file — to a section, paragraph, chunk, table, column or field — so that policy can permit part of an asset while restricting another part of the same asset.
What is element-level metadata?
A single contract may contain a public description, a commercially sensitive pricing schedule and regulated personal data. Classifying the file at its highest sensitivity makes it unusable for AI; classifying it at its lowest makes it unsafe. Element-level metadata resolves this by classifying constituent parts, so retrieval can return the permitted section and withhold the rest. Because AI retrieves fragments rather than documents, element-level granularity is the natural unit of governance for AI, and it is what makes fine-grained access control enforceable in practice.