As we speak, we’re introducing the Databricks AI Governance Framework (DAGF v1.0), a structured and sensible method to governing AI adoption throughout the enterprise.
As organizations embrace AI at scale, the necessity for formal governance grows. Enterprises should align AI improvement with enterprise targets, meet authorized obligations, and account for moral dangers. This framework is designed to assist program improvement, deployment, and steady enchancment.
The DAGF enhances the Databricks AI Safety Framework, providing an entire view of governance that spans each safety and operational integrity.
Why AI governance can’t wait
Based on a 2024 international survey of 1,100 expertise executives and engineers carried out by Economist Influence, 40% of respondents believed that their group’s AI Governance program was inadequate in guaranteeing the protection and compliance of their AI property and use circumstances. As well as, knowledge privateness and safety breaches had been the highest concern for 53% of enterprise architects, whereas safety and governance are essentially the most difficult features of information engineering for engineers.
As well as, in line with Gartner, AI belief, danger, and safety administration is the #1 prime technique development in 2024 that can issue into enterprise and expertise choices, and by 2026, AI fashions from organizations that operationalize AI transparency, belief, and safety will obtain 50% enhance by way of adoption, enterprise targets, and consumer acceptance.
Whereas it’s evident that the dearth of enterprise-level AI governance applications is quick turning into a key blocker to realizing return on worth from AI investments and AI adoption as a complete, we realized that there’s not a single, complete steerage framework that enterprises can leverage to construct efficient AI governance applications.
The 5 foundational pillars
On this framework, we introduce 43 key issues which can be important for each enterprise to grasp (and implement as acceptable) to successfully govern their AI journeys.
These key issues had been then logically grouped throughout 5 foundational pillars, designed and sequenced to replicate typical enterprise org-structures and personas.
Pillar I: AI Group
The AI Group pillar embeds AI governance throughout the group’s broader governance technique. It underscores the muse for an efficient AI program by means of greatest practices like clearly outlined enterprise targets and integrating the suitable governance practices that oversee the group’s individuals, processes, expertise, and knowledge. It explains how organizations can set up the oversight required to attain their strategic targets whereas lowering danger.
Pillar II: Authorized and Regulatory Compliance
The Authorized and Regulatory Compliance pillar helps organizations align AI initiatives with relevant legal guidelines and laws. It guides managing authorized dangers, deciphering sector-specific necessities, and adapting compliance methods in response to evolving regulatory landscapes. The result is AI applications are developed and deployed inside a sturdy authorized and regulatory framework.
Pillar III: Ethics, Transparency and Interpretability
The Ethics, Transparency, and Interpretability pillar helps organizations in constructing reliable and accountable AI methods. It emphasizes adherence to moral rules equivalent to equity, accountability, and human oversight whereas selling explainability and stakeholder engagement. This pillar offers strategies to ascertain accountability and construction inside organizational groups, serving to to make sure that AI choices are interpretable, aligned with evolving moral requirements, and fostering long-term belief and societal acceptance.
Pillar IV: Knowledge, AI Ops, and Infrastructure
The Knowledge, AI Operations (AIOps), and Infrastructure pillar defines the muse that helps organizations in absolutely deploying and sustaining AI. It offers pointers for making a scalable and dependable AI infrastructure, managing the machine studying lifecycle, and guaranteeing knowledge high quality, safety, and compliance. This pillar additionally emphasizes greatest practices for AI operations, together with mannequin coaching, analysis, deployment, and monitoring, so AI methods are dependable, environment friendly, and aligned with enterprise targets.
Pillar V: AI Safety
The AI Safety pillar introduces the Databricks AI Safety Framework (DASF), a complete framework for understanding and mitigating safety dangers throughout the AI lifecycle. It covers crucial areas equivalent to knowledge safety, mannequin administration, safe mannequin serving, and the implementation of sturdy cybersecurity measures to guard AI property.
For a further overview of DAGF and for an instance walkthrough of how a corporation can leverage the framework to create clear possession and alignment throughout the AI program lifecycle, please watch this presentation from the authors made in the course of the 2025 Knowledge + AI Summit.
Why Databricks is main this effort
As an trade chief within the knowledge and AI house, with over 15,000 prospects throughout numerous geographies and market segments, Databricks has continued to ship on its dedication to rules of accountable improvement and open supply innovation. We’ve upheld these commitments by means of our:
- Engagement with each trade and authorities efforts to advertise innovation and advocate for using secure and reliable AI
- Interactive workshops to coach organizations on how one can efficiently shepherd their AI journey in a risk-conscious method
- Open sourcing of key governance improvements equivalent to MLFlow and Unity Catalog, the trade’s solely unified answer for knowledge and AI governance throughout clouds, knowledge codecs and knowledge platforms.
These applications have supplied us distinctive visibility into sensible issues that enterprises and regulators face at the moment in AI governance. In furthering our dedication to serving to each enterprise succeed and speed up their Knowledge and AI journey, we determined to leverage this visibility to construct (and make freely accessible) a complete, structured and actionable AI Governance Framework.
Obtain the Databricks AI Governance Framework at the moment!
The Databricks AI Governance Framework whitepaper is now accessible for obtain. Please attain out to us through e mail at [email protected] for any questions or suggestions. Should you’re fascinated with contributing to future updates of this framework (and different upcoming artifacts) by becoming a member of our reviewer neighborhood, we’d love to listen to from you as nicely!
