Claris now supports the RBI FREE-AI Framework — full automated compliance mappingLearn More
Regulatory10 min read

RBI FREE-AI Compliance Checklist: What Every Indian Bank Needs to Audit

NA

Nodex8 AI Research

AI Research Team

February 28, 2026

AI Snapshot

3 things to know before you read

1

RBI's FREE-AI framework (Fairness, Resilience, Explainability, Ethics) is an audit-ready standard with specific technical requirements — not a set of aspirational principles

2

Indian banks must be able to produce feature-level explanations for any AI-driven credit decision within 24 hours of a customer request or regulatory inquiry

3

Resilience under FREE-AI requires not just uptime monitoring but continuous drift detection — banks must demonstrate that model logic has not materially shifted since last validation

What Exactly Does the RBI FREE-AI Framework Require?

Direct Answer

The RBI FREE-AI framework requires that AI systems used in regulated financial services demonstrate four properties: Fairness (no discriminatory outcomes across protected groups), Resilience (stability under data and distribution shifts), Explainability (human-readable rationale for decisions), and Ethics (human oversight and accountability). Each property has specific technical requirements that map to audit checkpoints.

The Reserve Bank of India released its FREE-AI guidelines as part of its broader responsible innovation framework. Unlike many regulatory frameworks that are principle-based and leave implementation open to interpretation, FREE-AI includes specific technical indicators that examiners look for. **F — Fairness:** - Demographic parity testing across religion, gender, geography, and employment type - Disparate impact analysis (the "4/5ths rule" adapted for Indian protected classes) - Feature Shielding documentation showing proxies for protected attributes have been identified and controlled - Quarterly fairness audits with documented remediation for any detected disparities **R — Resilience:** - Population Stability Index (PSI) monitoring at a minimum monthly cadence - Model Stability Score (MSS) tracking with defined drift thresholds and escalation procedures - Business Continuity Plan specifically addressing AI model degradation scenarios - Evidence of drift detection capability demonstrated during validation **E — Explainability:** - SHAP or equivalent feature attribution for every credit decision within 24 hours - Plain-language adverse action notices that translate feature attributions into customer-readable language - Audit trail of all explanations generated, stored for a minimum of 3 years - Documentation of XAI method selection and validation **AI — Ethics (Accountability + Integrity):** - Model governance register with version history, validation records, and change approvals - Named human accountability for each AI model in production - Ethics review committee sign-off for any new AI model deployment - Incident response procedure for model failures with defined SLAs
Full Article Access

Continue reading

Enter your details to unlock the remaining 2 sections — free access, no obligation.

Free · No spam · Unsubscribe any time

NA

Written by

Nodex8 AI Research

AI Research Team

The research team of Nodex8 AI focuses on global AI governance agenda, policy to code maturity across the globe, theoretical and empirical explainable AI research and technology advancement in the domain.

View research publications