Back to Trust Center
Ref 2d
AI Risk Register and Review Workflow
This document explains how AI-related risks are expected to be recorded, assessed, assigned, and reviewed before and after release.
Purpose
Show how AI risks are identified, scored, assigned, tracked, and closed before release.
Scope
- Risks involving safety, bias, privacy, reliability, and misleading AI behavior.
- Pre-release reviews and periodic re-evaluation after deployment.
- Ownership and tracking expectations for open AI risks.
Published Commitments
- Material AI risks should be recorded with severity, owner, status, and review date.
- Open high-priority risks should be reviewed before production release.
- Risk records should be updated when controls, integrations, or intended use change materially.