Risk taxonomy data model

[adhit-r]

Parent Issue

Part of #149 — AI Risk: IBM/MIT Risk Template Integration

Description

Define the risk category schema based on IBM AI Risk Atlas and MIT AI Risk Repository. Create database models for risk categories, risk factors, and assessment templates.

Requirements

• Risk category schema covering: fairness, explainability, robustness, privacy, security, accountability, hallucination
• Risk factors per category (sub-dimensions)
• Assessment template model — stores questionnaire structure per taxonomy
• Schema must be framework-agnostic so new taxonomies (OECD, custom) can be added without a redesign
• Follow existing SQLAlchemy model patterns in apps/backend/src/infrastructure/db/database/models.py

Deliverables

• Database models for risk categories and risk factors
• Assessment template model
• Alembic migration
• Seed data for initial IBM/MIT taxonomy
• Basic Pydantic schemas for the new models

Getting Started

1. Read apps/backend/src/infrastructure/db/database/models.py to understand existing model patterns
2. Read apps/backend/src/domain/compliance/ for how compliance models are structured — risk assessment should follow similar patterns
3. Create models, migration, and seed data
4. Open a PR against main and link this issue

[IlyasFardaouix]

I've taken a closer look at apps/backend/src/infrastructure/db/database/models.py and it seems like we're already following SQLAlchemy model patterns there. For the risk category schema, I'm wondering if we could leverage the existing Category model as a base class and extend it with the specific risk categories (fairness, explainability, etc.)? This would help us keep the schema framework-agnostic and make it easier to add new taxonomies like OECD or custom ones. Should we discuss this approach further or do you have any thoughts on how to proceed?