What It Is
The MRIS (Model Residual-Risk Information Sheet) is a disclosure template for documenting what risks remain after an AI model has been tested, validated, and deployed. Think of it as a bill of lading for AI models — a structured handoff document that tells downstream users what they’re inheriting.
When to Use
Use the MRIS when:
- Deploying a model to production
- Handing off a model to another team or organization
- Documenting model risks for regulatory compliance
- Creating model cards or system cards
How to Use
- Complete each section before deployment
- Review with your risk/compliance team
- Include the MRIS in your model documentation
- Update when model behavior changes
- Share with downstream users and stakeholders
Inline Preview
| Section | Content |
|---|---|
| Model Name | [Name and version] |
| Purpose | [What this model does] |
| Training Data | [Data sources and date ranges] |
| Known Limitations | [What the model cannot do reliably] |
| Bias Risks | [Known or suspected biases] |
| Edge Cases | [Scenarios where performance degrades] |
| Monitoring Requirements | [What to watch for in production] |
| Escalation Contacts | [Who to contact for issues] |
| Version History | [Changes and when they were made] |