Fabric Notebooks: Using a Shared Redaction Config to Scrub PII Before Data Leaves the Lakehouse

A DEV Community article argues that data exported from a Fabric lakehouse should pass through one shared redaction configuration rather than ad hoc scrubbing in individual notebooks. Free-text columns in CSV dumps, partner shares, and ticket attachments are identified as the main places where names, phones, and emails tend to appear. The author recommends a single config cell or module defining enable flags, column lists, regex patterns, and replacement tokens, with publish paths blocked unless redaction runs. Sensitivity labels are described as useful for classification and access control but not as a substitute for rewriting exported data. The piece also suggests recording residual-risk decisions explicitly and testing redaction with fixtures.