databricks Databricks Blog ·

Databricks expands data governance for AI beyond security

bloggovernancedatabricksarchitect
security

Databricks introduces a comprehensive vision for data governance, emphasizing knowledge, context, and ontology in addition to traditional security measures. This approach, part of their Data Empowerment Program (DEP), aims to transform compliance artifacts into a foundation for building trusted and cost-effective AI models. The framework leverages Unity Catalog as a runtime for automated 'agents' that manage the data and AI lifecycle, shifting from pipeline-centric to context-centric engineering. This new governance model seeks to enable AI systems with higher trust and better economics by using governed data semantics.

Security (1)
  • Security / IAM: Owns the classification tiers and access attributes that automatically drive de-identification and row-level entitlements. The lifecycle we described has a hard prerequisite hiding inside it: every one of those test and evaluation stages needs realistic data to run against – and in healthcare, you can’t test real PHI. So, the challenge becomes the need for realistic test data everywhere without compromising security. De-identification is how we keep data analytically useful and safe. Where does the de-identification agent get its knowledge? Not from a hand-maintained spreadshee

Read the original announcement →

https://www.databricks.com/blog/governance-beyond-security-knowledge-context-ontology-lakehouse

Related releases