# Looker restricts a sensitive field through access rules, not labels, and keeps 90 days of query history

GUIDE, read 13 Nov 2026

**The short answer.** Looker's LookML tags are not used by the Looker model, and hidden is not a security feature, per Google Cloud's docs. Access grants, access filters and sql_always_where do the restricting, and System Activity keeps the History table for 90 days. Atlan, Alation, DataHub, Select Star, Secoda and Collibra document Looker.

**Where Euno is behind.** Euno is not a policy engine or a DSPM: it reads metadata only, so it does not scan data contents or mask at the warehouse, which leaves it behind Atlan on acting on findings (5 to 7), and it publishes no dollar prices.

## Looker marks fields with tags that its own model ignores

LookML lets a developer add a tags parameter to a dimension or measure. Google Cloud's docs say these strings are not used by the Looker model. They can be passed to other applications through integrated services on the Action panel or through API calls. A tag named pii therefore describes a field to outside software and enforces nothing in Looker.

The hidden parameter is a similar trap. The docs say a hidden field is still accessible in the UI if someone adds it to the URL, and that hidden keeps the field picker clean and is not a security feature. A governance review that counts hidden fields as protected fields is counting the wrong thing.

## Access grants, user attributes and filters do the restricting

An access grant is defined in a model file against a user attribute and a list of allowed values. Adding required_access_grants to an Explore, join, view or field limits it to users who hold every listed grant. Users without them do not see the field in the picker, and a Look that uses it shows a warning that the field no longer exists or is not accessible.

An access_filter applies a user attribute as a filter on a field, per Explore. Google's page is direct about the failure mode: if you forget to add access_filter to an Explore that should have it, the data will not be restricted and users can see all of it in that Explore. The sql_always_where parameter inserts a condition into the WHERE clause that users cannot change, applies to dashboards, scheduled Looks and embedded content on that Explore, and is not shown unless the user reads the generated SQL.

Content access is a separate setting. It decides who may view or manage a folder, and the admin docs say data access, feature access and content access combine to set what a user can do. A field protected by an access grant can still sit in a folder many people can open, so review the three together.

## The warehouse sees Looker's identity, not always the viewer's

Looker connects to BigQuery either with a service account or with per-user Google OAuth. Looker's BigQuery page says OAuth lets database administrators audit which Looker users are running queries and enforce role-based controls with Google permissions. With a service account, BigQuery sees one principal for everyone. BigQuery column-level security with policy tags is checked against the identity that runs the query, so policy tags protect a Looker user's view only where that user's own identity reaches BigQuery.

A second BigQuery rule matters for rebuilt tables. BigQuery's docs say a write with the WRITE_TRUNCATE disposition removes existing policy tags unless the destination schema supplies them, while WRITE_APPEND preserves them. Ask the Looker team whether any table a model reads is overwritten that way.

## System Activity answers who ran what, for a limited window

System Activity gives admins Explores over Looker's internal database. The usage docs state that the History table is truncated to the past 90 days, that most tables are truncated on a schedule, and that Elite System Activity is the route to longer retention. The Explore refreshes every 12 hours. The History Explore covers only content executed within the retention window, so a dashboard nobody opened recently appears in the Dashboard Explore and not in History.

One detail is easy to miss and affects audit work. The Role view reflects current role assignments, so joining History to Role shows a user's current role next to their past queries, not the role they held when the query ran. An auditor asking who could see a field last quarter gets an answer that may be wrong for anyone whose role changed. Export History on a schedule and store the role snapshot beside it.

## Ranked tools that document Looker, and what each adds

Atlan documents field-level lineage for Looker views, explores, looks, tiles and dashboards. DataHub lists column-level lineage for Looker as enabled by default and can pull historical dashboard usage from Looker System Activity, which is the 90-day History data described above. Select Star says its Looker integration gives column-level lineage from source views to LookML views, explores and dashboards. Alation, Collibra, Secoda and Euno also list Looker among the BI tools they read. None of these descriptions says the tool reads LookML access grants or access filters; none found on 13 Nov 2026 in the evidence files for this site, so ask each vendor whether the lineage view shows which fields are restricted.

On the exposure map, the BI field stage reads partial for Atlan, documented for Alation, partial for DataHub, documented for Select Star, partial for Secoda, partial for Collibra and documented for Euno. The report stage reads partial, documented, partial, documented, partial, partial and documented in the same order. See [the BI field column](/exposure-map/bi-field/) for the rows behind each cell.

Take the field-level view from the tools above, not the restriction view: the restriction is in LookML and the connection.

## Sign-off for Looker comes down to four checks

Before sign-off, list the explores that expose a sensitive source column, confirm each has an access_filter or a required_access_grants entry, confirm the connection uses per-user OAuth where policy tags are the control, and schedule the History export before the 90-day window drops the evidence.

## Sources

- Google Cloud: LookML field tags: https://cloud.google.com/looker/docs/reference/param-field-tags (read 13 Nov 2026)
- Google Cloud: LookML field hidden: https://cloud.google.com/looker/docs/reference/param-field-hidden (read 13 Nov 2026)
- Google Cloud: required_access_grants for fields: https://cloud.google.com/looker/docs/reference/param-field-required-access-grants (read 13 Nov 2026)
- Google Cloud: LookML access_filter: https://cloud.google.com/looker/docs/reference/param-explore-access-filter (read 13 Nov 2026)
- Google Cloud: LookML sql_always_where: https://cloud.google.com/looker/docs/reference/param-explore-sql-always-where (read 13 Nov 2026)
- Google Cloud: Looker access control: https://cloud.google.com/looker/docs/access-control-and-permission-management (read 13 Nov 2026)
- Google Cloud: Looker BigQuery connection: https://cloud.google.com/looker/docs/db-config-google-bigquery (read 13 Nov 2026)
- Google Cloud: BigQuery column-level security: https://cloud.google.com/bigquery/docs/column-level-security-intro (read 13 Nov 2026)
- Google Cloud: System Activity Explores: https://cloud.google.com/looker/docs/usage-reports-with-system-activity-explores (read 13 Nov 2026)
- Atlan: what Atlan crawls from Looker: https://docs.atlan.com/llms/connectors/looker/what-does-atlan-crawl-from-looker/llms.txt (read 13 Nov 2026)
- DataHub: Looker source: https://docs.datahub.com/docs/generated/ingestion/sources/looker (read 13 Nov 2026)
- Select Star: Looker integration: https://www.selectstar.com/integrations/looker (read 13 Nov 2026)

Last reviewed 13 Nov 2026. Canonical: https://aidatagovernancetools.com/guides/sensitive-fields-in-looker/
