They thought the query was gone. It wasn’t. It was just hiding in the dark corners of a warehouse of tables no one knew existed.
This is the hidden cost of poor discoverability in Databricks: you don’t know what’s there, so you can’t protect it. Access control without discoverability is a locked door in a maze with no map. It doesn’t matter how strong the lock is if you can’t even find the door.
Databricks has powerful tools for securing data: Unity Catalog, fine-grained ACLs, and role-based permissions. They let you define who can see, query, and edit objects. But there’s a deeper problem. Engineers often set permissions based on the resources they know exist. Many critical assets never enter their view. This leads to sensitive data sitting in plain sight, ungoverned.
Good access control begins with complete visibility. You need a real index of every table, schema, and notebook across all workspaces. You need to see data lineage to understand where sensitive columns flow. You need auditing that confirms nothing is slipping past the rules.