Governing AI Agents
by DeepLearning.AI and Databricks
Build an AI agent that touches sensitive data safely: catalog permissions, agent identity, data masking and MLflow deployment on Databricks.
Overview
Governing AI Agents is a beginner-level DeepLearning.AI short course built in collaboration with Databricks and taught by Amber Roberts, a Staff Technical Marketing Manager at Databricks; DeepLearning.AI announced it on October 24, 2025. It takes about 1 hour 30 minutes and is organised as 9 video lessons plus a graded quiz, with roughly half of that runtime spent in three hands-on labs. The conceptual lessons define AI agent governance through four pillars, namely lifecycle management, risk management, security and observability, then show how to establish governance in a data catalog and how to add an identity to an agent. Lab 1 builds the governance foundation in Databricks, where the course pairs Unity Catalog for organising data and permissions with agent identity and access management and data masking. A lesson on building, evaluating and deploying agents leads into Lab 2, which builds an HR analytics agent on that governed data, and Lab 3, which evaluates the agent and deploys it with MLflow. The motivating risk, in the course announcement's own words, is concrete: without proper controls an agent can accidentally expose personal information, modify sensitive records or operate beyond its intended scope. The trade-off is that everything is taught inside Databricks, so the patterns of least-privilege access for agent identities, masking sensitive fields and monitoring deployed agents transfer to other platforms, but the specific steps do not. A learner's public companion repository notes that the labs run on a free Databricks workspace, and the course page lists access as free during DeepLearning.AI's limited-time platform beta.
At a Glance
- Topic
- Agentic
- Level
- Beginner
- Format
- Course
- Cost
- Free
- Duration
- ~1.5 hours, self-paced (9 video lessons including 3 labs, plus a graded quiz)
- Provider
- DeepLearning.AI and Databricks
- Hands-on
- Yes — code/exercises
- Certificate
- None
What You’ll Learn
- ✓Explain AI agent governance through lifecycle management, risk management, security and observability pillars
- ✓Organise agent-accessible data and permissions in Databricks Unity Catalog as a governance foundation
- ✓Give an AI agent its own identity and manage what data that identity can access
- ✓Implement data masking so sensitive personal fields are hidden from the agent
- ✓Build an HR analytics agent on top of governed Databricks data
- ✓Evaluate the HR analytics agent and deploy it to production with MLflow
- ✓Recognise how ungoverned agents expose personal information, modify sensitive records or exceed their scope
Highlights
- •Treats agent governance as an engineering task with hands-on labs rather than a policy overview
- •About half of the 90-minute runtime is spent in three hands-on Databricks labs
- •Uses a realistic sensitive-data scenario, an HR analytics agent, where access control and masking obviously matter
- •Covers agent identity as a first-class control, which most agent-building courses never mention
Who It’s For
Best For
- ✓AI engineers deploying agents that touch personal or regulated data
- ✓Data platform teams on Databricks adding agents to an existing lakehouse
- ✓Security and governance leads who want a hands-on view of agent access controls
Prerequisites
- •Basic Python familiarity
- •A free Databricks workspace account to complete the labs
FAQ
What is Governing AI Agents?
A beginner DeepLearning.AI short course, built with Databricks, for AI builders whose agents access sensitive or personal data. You learn four governance pillars, set up permissions and data masking in Unity Catalog, give an agent its own identity, then build, evaluate and deploy an HR analytics agent with MLflow so it stays safe, compliant and observable in production.
Is Governing AI Agents free?
Governing AI Agents is free to access.
What level is Governing AI Agents for?
Governing AI Agents is aimed at a beginner audience. Recommended background: Basic Python familiarity, A free Databricks workspace account to complete the labs.
How long does Governing AI Agents take?
Expect roughly ~1.5 hours, self-paced (9 video lessons including 3 labs, plus a graded quiz). Most learners work through it at their own pace.
What will I learn from Governing AI Agents?
You'll learn: Explain AI agent governance through lifecycle management, risk management, security and observability pillars; Organise agent-accessible data and permissions in Databricks Unity Catalog as a governance foundation; Give an AI agent its own identity and manage what data that identity can access; Implement data masking so sensitive personal fields are hidden from the agent; Build an HR analytics agent on top of governed Databricks data; Evaluate the HR analytics agent and deploy it to production with MLflow; Recognise how ungoverned agents expose personal information, modify sensitive records or exceed their scope.
Topics
Sources
This page was written from 3 sources, 2 on domains other than deeplearning.ai.