Model Risk Assessment Framework
Designed a tiered risk assessment that every new AI feature passes through before launch.
- ORGANISATION
- Earthstorm Media
- ROLE
- AI Governance Lead
- DATE
- Jun 2025
Risk ManagementAssuranceEU AI ActNIST AI RMF
Overview
I built a questionnaire-driven framework that classifies AI use cases into risk tiers modelled on the EU AI Act. Each tier unlocks a proportionate set of controls, from a simple checklist for low-risk features to a full impact assessment and sign-off for high-risk ones.
Outcomes
- Risk tiering applied to 12 AI features in the first quarter
- Average review time cut from 3 weeks to 5 days
- Reusable templates for impact assessments and sign-off