September 2025
These features and Databricks platform improvements were released in September 2025.
Releases are staged. Your Databricks account might not be updated until a week or more after the initial release date.
MLflow metadata is now available in system tables (Public Preview)
September 5, 2025
MLflow metadata is now available in system tables. View metadata managed within the MLflow tracking service from the entire workspace in one central location, taking advantage of all the lakehouse tooling Databricks offers, such as building custom AI/BI dashboards, SQL alerts, and large scale data analytic queries.
For information, see MLflow system tables reference.
Databricks Assistant Agent Mode: Data Science Agent is in Beta
September 3, 2025
Agent Mode for Databricks Assistant is now in Beta. In Agent Mode, the Assistant can orchestrate multi-step workflows from a single prompt.
The Data Science Agent is custom-built for data science workflows and can build an entire notebook for tasks like EDA, forecasting, and machine learning from scratch. Using your prompt, it can plan a solution, retrieve relevant assets, run code, use cell outputs to improve results, fix errors automatically, and more.
See Use the Data Science Agent.
C5 compliance controls
September 2, 2025
C5 compliance controls provide enhancements that help you with compliance for your workspace. C5 is a German Federal Office for Information Security (BSI) standard that defines minimum security requirements for cloud service providers. See Cloud Computing Compliance Criteria Catalog (C5).
Tables backed by default storage can be Delta shared to any recipient (Beta)
September 2, 2025
Delta Sharing providers can now share tables backed by default storage with any recipient, including both open and Databricks recipients—even if the recipient is using classic compute. Tables with partitioning enabled are an exception.
See Limitations.
Migrate Lakeflow Declarative Pipelines from legacy publishing mode is rolled back to Public Preview
September 2, 2025
Lakeflow Declarative Pipelines includes a legacy publishing mode that previously limited publishing to a single catalog and schema. Default publishing mode enables publishing to multiple catalogs and schemas. A feature, recently released as generally available, can help migrate from the legacy publishing mode to the default publishing mode. Due to an issue found after release, the migration feature has been rolled back to Public Preview status and functionality.
See Enable the default publishing mode in a pipeline
AI agents: Authorize on-behalf-of-user Public Preview
September 2, 2025
AI agents deployed to Model Serving endpoints can use on-behalf-of-user authorization. This lets an agent act as the Databricks user who runs the query for added security and fine-grained access to sensitive data.
SQL Server connector supports SCD type 2
September 1, 2025
The Microsoft SQL Server connector in Lakeflow Connect now supports SCD type 2. This setting, known as history tracking or slowly changing dimensions (SCD), determines how to handle changes in your data over time. With history tracking off (SCD type 1), outdated records are overwritten as they're updated and deleted in the source. With history tracking on (SCD type 2), the connector maintains a history of those changes. See Enable history tracking (SCD type 2).