Embedded AI/BI Dashboards added viewer-level row filtering through signed tokens.

In September 2026, Databricks introduced viewer-specific row filtering for embedded AI/BI Dashboards [7]. The pattern used signed embed tokens to pass external viewer attributes into an entitlements table query. This design allowed a single published dashboard to restrict data visibility dynamically according to identity-provider group memberships, eliminating the need to duplicate dashboards or manually adjust individual query filters [7].

Community members focused heavily on the mechanics and overhead of row-level security across enterprise architectures. Practitioners compared the runtime query performance of native row filters directly against traditional authorized views on large datasets [4]. Discussions also explored design patterns for filtering rows when the required authorization attributes resided in external lookup tables rather than on the target table itself [3], alongside operational strategies for synchronizing row-level security policies across multiple catalogs that shared identical schema definitions [5].

Securing emerging AI workloads with row filters was another prominent theme. Engineers evaluated how Unity Catalog security rules applied to automated queries, clarifying what native row-level protections the platform enforced automatically versus what developers had to implement manually for retrieval-augmented generation agents [1]. Related community work addressed filtering and validation gates needed before exposing governed tables to Genie spaces [2].