Databricks announced the beta of FILE type, a new native column type for storing unstructured data—such as contracts, product images, call recordings, and video—as governed, queryable columns in tables. FILE enables unified governance by applying the same fine-grained access controls and security policies to raw files as to standard tables. It supports automated compliance, so deleting a row containing a FILE also deletes the file’s binary in object storage, facilitating GDPR "right to be forgotten." The type supports standard SQL and Python UDFs directly on unstructured files, and allows building materialized views to run AI functions incrementally. FILE columns store lightweight pointers rather than heavy file binaries, processing actual file content only when a query explicitly needs it. Databricks is working with the community to build support into Parquet, Delta Lake, Apache Iceberg, and Apache Spark. Use cases include enterprise document assistants, visual inspection over product images, call analytics over audio recordings, video understanding, and multimodal retrieval for agents. The announcement includes an example from an autonomous-driving company’s CEO asking whether random stops can be reduced, using dashcam videos stored as FILE columns.