Version: Cloud

v1.50

Release notes

Checklist templates are removed in favor of the controls as part of our new control framework feature.

Control frameworks

Take charge of your governance workflow with our new control frameworks!
- Govern your Deployments with control frameworks that translate regulations and policies into actionable controls
- Create tailored control frameworks or leverage our default Responsible AI control framework
- Apply frameworks to Workspaces where we automatically list relevant controls per use case, based on the use case details
- Centralize governance for multiple Deployments in one location with our use cases. By implementing use cases, you create a structured framework to organize, monitor, and control your AI systems throughout their entire lifecycle
- Use automated checks to track your progress towards a control and submit supporting evidence. Complete controls to track how your use case measures up against the framework
The new control frameworks feature helps you implement structured governance across your AI deployments with greater flexibility and oversight.
Guardrails
Implement regex-based guardrails to automatically remove sensitive information from model inputs and/or outputs
- Available for both managed and external deployments
- Choose from standard guardrails or write your own custom ones
Upload documents in PDF format to provide organization-wide policies or procedures.

Added support for submitting metadata without repository for external and registration Deployments
Renamed 'team' to 'organization'
Added strict validation on provided reference.json and metadata.json
Switching between having a linked repository and no repository has been made more robust
Improve ID accessibility (click to copy) in prediction logs