01
AI governance that stands up to a review
FINMA Guidance 08/2024 expects an inventory of all AI applications, a consistent risk classification, testing and ongoing monitoring, documentation and explainability. Material applications additionally require an independent review, separate from development. That is where I start: taking stock, assessing, and for the applications with obligations a review procedure an audit firm can follow.
- AI inventory and risk classification
- Model and data quality, stability, bias
- Explainability towards clients and the supervisor
- Independent review separate from development
- Assistants and automation on your own data
What you are left with
- 01An inventory that stands up to a review
- 02A review report with findings and measures, not just an opinion
- 03Clear rules on who sees which data and who signs off results
