RBI Model Risk Readiness Scorecard - Risk Management Association of India

Mandate 1 of 5

Kill Switch

The ability to instantly override or deactivate any AI model in use.

If a model started producing harmful outputs today, could you suspend it within minutes following a written process? Yes, documented and tested We have a process, but it has never been tested No formal process exists

Select an option to continue.

Is responsibility for triggering that override assigned to a named role? Yes, clearly assigned Informally understood, not documented Not assigned to anyone

Select an option to continue.

Mandate 2 of 5

Human Oversight

Documented, demonstrably robust human-in-the-loop checkpoints.

For AI-driven decisions that affect customers, is there a documented human checkpoint before the decision is finalised? Yes, for all relevant models For some models only No, decisions are automated end-to-end

Select an option to continue.

Could someone on your team explain, in plain language a regulator would accept, why a specific model produced a specific output? Yes, consistently Sometimes, depending on the model Not reliably

Select an option to continue.

Mandate 3 of 5

Customer Disclosure

Telling customers when AI drives a decision about them.

Do customers know when an AI model has materially influenced a decision about them? Yes, consistently disclosed Disclosed in some products only Not currently disclosed

Select an option to continue.

Is that disclosure a standard, repeatable process rather than handled case by case? Yes, standardised Partially standardised No, ad hoc

Select an option to continue.

Mandate 4 of 5

Vendor Risk

Actively managing third-party AI providers, not just using them.

Do your contracts with third-party AI vendors include audit, override, and disclosure rights? Yes, in all relevant contracts In some contracts Not currently

Select an option to continue.

Could you produce a complete inventory of every AI/ML model in use, including vendor-supplied ones, on demand? Yes, today Partially, with some gaps No, not currently

Select an option to continue.

Mandate 5 of 5

Board Accountability

Governance owned at the Board, not delegated entirely to tech teams.

Does your Board receive structured reporting on AI model risk, separate from general IT updates? Yes, regularly Occasionally No

Select an option to continue.

Could your Board answer right now: "What would we shut down, and how fast?" Yes, with confidence Possibly, but it has not been tested No

Select an option to continue.

Your Readiness Result

0 / 20

Governance & Framework

Operational & Vendor Risk

Think your risk team should take this assessment?

Download the full compliance checklist

Get the complete RBI AI Model Risk Management Compliance Checklist with 40+ specific action items, a priority matrix, and implementation timeline. Fill in your details to download.

We will not spam you. Your details unlock the downloadable checklist.

Want the full regulatory breakdown? Read RBI's draft guidance explained. Comments close 24 July 2026.

Is your AI governance actually regulator-ready?

Kill Switch

Human Oversight

Customer Disclosure

Vendor Risk

Board Accountability

Download the full compliance checklist