Databricks details Omnigent policies for governing AI agents

Databricks detailed the governance capabilities of Omnigent policies for AI agents. The policies inspect proposed actions before execution to control cost and behavior, allowing users to configure spending limits and restrict tool usage.

The system also tracks accumulated risk across an entire session, enabling actions to require approval once a predefined threshold is reached. Further details and implementation examples are available in Databricks' blog post and the accompanying GitHub repository.