Hong Kong AI Risk Gating Criteria
Tags: Frameworks
Purpose
Risk Gating Criteria are used within the AI Application Impact Assessment (AAIA) to categorise AI projects as high‑risk or non‑high‑risk, setting the depth of assessment, escalation pathways, and senior oversight requirements.
How It Works
- Applied at the Project Strategy stage to determine the level of assessment needed.
- Initial AAIA responses inform risk status and the scope of subsequent questions and reviews.
- High‑risk classification triggers enhanced governance (e.g., IT Board/CIO review) beyond PSC/PAT endorsement.
Governance and Approvals
- Non‑high‑risk: PSC/PAT review and endorsement after initial assessment.
- High‑risk: PSC/PAT review plus IT Board/CIO review and endorsement, with ongoing monitoring obligations.
Integration with Lifecycle Reviews
- Risk status is re‑evaluated at key lifecycle checkpoints (planning, development, pre‑rollout, and post‑release) with updates to prior answers to reflect current positions.
- Procurement of third‑party technology or data requires completion/review of dedicated questions before procurement decisions.
Benefits
- Ensures proportional oversight, focusing senior attention and independent challenge on applications with greater potential impact.
- Provides a consistent, documented method for triaging projects and aligning controls with the framework’s Ethical AI Principles.
Sources:- Ethical_AI_Framework_en.pdf
← Back to Knowledge Base