Courts analytics must ensure each clerk office sees only cases for its own jurisdiction code in a shared Unity Catalog table. What should they implement?

Row-level security restricts which rows each principal sees in a shared Unity Catalog table. Courts analytics should implement RLS so each clerk office only returns cases matching its jurisdiction code, even though all jurisdictions share one governed table. That is the clerk-counter rule: one cabinet, filtered dockets—preventing cross-jurisdiction peeking that would be a privacy and ethics failure.

Applying Unity Catalog column masks so each clerk office only sees masked jurisdiction code values only masks that code value but still returns rows from other jurisdictions; masks hide fields, not rows. Provisioning separate Unity Catalog schemas per jurisdiction code so clerk offices query isolated case tables fragments the shared table into many isolated tables and adds maintenance overhead when RLS can enforce tenancy inside one table. Adding Unity Catalog CHECK constraints on jurisdiction code so clerk offices insert only valid case rows only validates values on write and does not filter reads, so offices would still see other jurisdictions' cases.

Exam caveat: RLS policies must be correct and tested for both allow and deny paths; miswritten policies can over-expose or empty results. Do not confuse RLS (rows) with column masks (fields) or separate schemas (objects) or CHECK constraints (insert validation). Operational check: define policies mapping clerk groups to jurisdiction codes, query as each office, confirm only matching cases return, attempt cross-jurisdiction filters and expect empty/unauthorized rows, and include RLS tests in pipeline CI before expanding clerk onboarding.