Databricks added token-driven row-level security controls for embedded AI/BI Dashboards.

In September 2026, the primary platform update focused on access control for embedded AI/BI Dashboards [7]. Databricks detailed an architecture that allows a single embedded dashboard to serve multiple viewers securely by combining an entitlements table with signed embed tokens using the __aibi_external_value parameter [7]. This approach enforced row-level data access directly through identity provider group memberships, which eliminated the need for administrators to build separate dashboards or maintain repetitive filter logic across individual queries [7].

Meanwhile, community activity focused heavily on advanced design techniques and visual customization. Practitioners published step-by-step guides for building dumbbell charts [9], using custom HTML to format table widgets [3], and configuring dynamic chart titles that update based on user selections [2]. These practical tutorials ran alongside broader debates evaluating how well Databricks AI/BI Dashboards compete against established enterprise reporting tools such as Power BI and Tableau [1].

Practitioners also flagged functional limitations in everyday dashboard operations. For example, community members noted that exporting data to CSV from an AI/BI Dashboard table widget failed to preserve the visual column order established in the dashboard layout, instead reverting to the order returned by the underlying query [6].