Insights / RLS in Power BI / Part 8

Part 8 of 20 · Build it

OLS vs CLS vs data masking in Power BI and Fabric

By Zohal Zahir · November 2026 · From the series Who sees what: RLS in Power BI

RLS hides rows. To hide a column (salary, email, card number) there are three tools. Which one you need depends on how the person reaches the data.

First, one rule for all of them: they only apply to Viewers. Admin, Member and Contributor see everything.

There are two doors to your data. Security only works on the door the person actually walks through.

Door 1: the semantic model. Guarded by RLS (rows) and OLS (columns). People come this way when they open a report or app, build their own report on the model, or use Analyze in Excel.

Door 2: the source (SQL analytics endpoint, Warehouse). Guarded by CLS and dynamic data masking (DDM). People come this way with direct SQL, notebooks, other models built on the same warehouse, other tools, and Direct Lake with the user's own identity. OLS isn't on this path.

Peoplereports, Excel,SQL, notebooksDoor 1: semantic modelRLS hides rowsOLS hides columnsDoor 2: the sourceSQL endpoint / WarehouseCLS blocks · DDM masksYour datasalary, email, IDsreport · app · ExcelSQL · notebooks · toolsSecurity only works on the door the person walks through.
Tool Where What it does
OLS Semantic model, per role Hides the column completely. Visuals that use it show an error
CLS Warehouse / SQL endpoint Blocks the column at the source
DDM Warehouse / SQL endpoint Shows the column with masked values
role 'Hide salary'
    modelPermission: read
    tablePermission Employees
        columnPermission Salary = none
-- CLS: GRANT is a whitelist. Listed columns show; the rest are blocked
GRANT SELECT ON Customers(Name, Email) TO [Sales Group];
-- or, if they already have SELECT on the table
DENY SELECT ON Customers(CreditCard) TO [Sales Group];

-- DDM: keep the column, mask the values
ALTER TABLE Customers ALTER COLUMN Email ADD MASKED WITH (FUNCTION = 'email()');
GRANT UNMASK ON Customers TO [Finance Group];

Gotchas. DDM is not a security boundary: people can guess values with range queries, so pair it with CLS or RLS. Source security only tells users apart if their own identity reaches the source (Part 9). With Direct Lake on the SQL endpoint, SQL security changes how queries run, so test visuals that use protected columns.

Check before go-live

  • Sensitive columns are hidden with OLS in the model Desktop
  • Sensitive columns are also protected at the source if people can query it directly
  • The row filter and the OLS rule for an audience sit in the same role (Part 14) Desktop

Want a second pair of eyes on your own model? The RLS review checks your roles, relationships and workspace access against this checklist, at a fixed price.

Found this useful? Share it.Tag Zohal Zahir on LinkedIn so she can say thanks.

Want your RLS checked on a real model?

Book a free 30-minute call. Tell us how your model and workspaces are set up, and we'll tell you where access could leak and what fixing it would involve.

Book a free call30 minutes. Reply within two business days