Quick answer
Economic Survey 2025-26 reports AI adoption or development at about 21% of surveyed financial institutions; RBI’s FREE-AI pairs innovation with safeguards.
AI in financial services uses data-driven models to support tasks such as credit assessment, fraud detection and customer service. Box III.2 argues that innovation and risk management are complements: inclusion gains require trustworthy data, explainable decisions and accountable institutions.
Concept diagram
An AI teller needs a human risk shield
AI can improve financial services while creating new governance challenges.
- AI can aid customer support, fraud detection and credit analysis.
- A useful model is not automatically fair, secure or explainable; oversight remains necessary.
Basics: what changes when finance uses AI
Alternative credit scoring
- Transaction histories can supplement conventional collateral and bureau records for thin-file borrowers.
- A predictive relationship is not proof of causation; biased historical data can reproduce exclusion.
Human accountability
- A financial institution remains responsible for model-assisted decisions.
- Explainability means making the basis of a decision understandable; assurance means testing and monitoring whether the system remains reliable.
Box III.2: opportunity, but an uneven starting point
Early-stage domestic adoption
- Approximately 21% of surveyed banks and financial institutions were implementing or developing AI solutions.
- Larger banks lead; smaller urban cooperative banks and many NBFCs face data, talent and IT-budget constraints.
Global outlook is a projection
- The WEF white paper cited by the Survey projects financial-services AI investment of US$97 billion by 2027.
- Typical Indian uses remain process efficiency, simple chatbots, lead generation and decision support—not widespread autonomous decisions.
Data chart
Institutions implementing or developing AI: surveyed Indian sample
Source: Economic Survey 2025-26, Box III.2 (RBI FREE-AI Committee Report) · Approximately 21%; this is a survey finding, not a census of all institutions.
Show the data (Per cent)
| Period / item | Share of surveyed institutions |
|---|---|
| Implementing or developing AI | 21 |
FREE-AI: enable innovation and contain risk
Enable
Infrastructure, Policy and Capacity Building: shared data/technology, agile sandboxes and skills.
Protect
Governance, Protection and Assurance: board oversight, privacy/security, validation and continuous monitoring.
Align
Support India AI Mission capabilities and consistency with the Digital Personal Data Protection Act.
Seven Sutras: translate principles into institutional practice
| Principle(s) | Operational implication |
|---|---|
| Trust; People First | Protect consumer confidence and place people’s welfare above model convenience. |
| Innovation over restraint | Permit controlled experimentation rather than reflexively prohibit new tools. |
| Fairness and equity; Accountability | Test exclusionary outcomes and assign responsibility for decisions. |
| Understandable by design | Build explanations into systems rather than add them after a complaint. |
| Safety, resilience and sustainability | Test failures, maintain fallbacks and monitor changing model behaviour. |
Exam traps: keep the distinctions clear
Statements examiners like to twist — learn the correct version.
FREE-AI is only a ban-oriented framework.
Its enablement and risk-mitigation pillars deliberately operate together.
21% means AI is making autonomous lending decisions everywhere.
The estimate includes development-stage solutions and mostly basic applications.
US$97 billion is India’s realised AI spending.
It is a global financial-services investment projection for 2027 cited from WEF.
Prelims focus
- Recall the seven Sutras and the six pillars; distinguish principles from implementation pillars.
- ECB: statistics and inflation nowcasting; Bank of Canada: inflation, activity and banknote demand.
Mains angle
- Assess whether shared AI infrastructure can reduce the technology gap between large banks and smaller lenders.
- Discuss a proportionate approach combining sandboxes, explainability, data protection and board accountability.
Flip & recall
Answer in your head first, then tap the card to check.
Practice MCQs
Original Prelims-style questions with explanations.
0 / 3 correct · 3 left
Q1.Which is a risk-mitigation pillar of FREE-AI?
Q2.The 21% figure in Box III.2 refers to:
Q3.Which reading of the WEF estimate is correct?
