Quick summary
AI-powered lending: Artificial intelligence can improve credit assessment, reduce processing time and support financial inclusion, but biased data may produce unfair outcomes.
Fraud detection: AI can identify suspicious transactions and emerging fraud patterns, although false positives remain a concern.
Operational efficiency: Automation can streamline document processing, compliance, customer service and internal banking operations.
Data privacy: AI systems require careful management of sensitive financial information, consent, access controls and third-party data sharing.
Cybersecurity: Model manipulation, prompt injection, data leakage and technology dependence introduce new security risks.
Human oversight: Banks must retain meaningful control over consequential automated decisions.
Expert outlook: Sustainable AI adoption requires strong governance, independent model validation, responsible innovation and customer trust.
An Expert Perspective on Artificial Intelligence, Digital Lending, Cybersecurity and Responsible Automation
Artificial intelligence is rapidly becoming one of the most consequential technologies affecting modern banking.
For decades, banks have relied on software systems to process transactions, maintain customer records, calculate interest, manage accounts and monitor financial activity.
Artificial intelligence introduces a different level of capability.
Instead of simply executing predefined instructions, modern AI systems can identify patterns in large datasets, generate predictions, analyse documents, detect unusual activity and assist with decisions that previously required substantial human effort.
For India's banking industry, these capabilities create significant opportunities.
AI could improve credit appraisal, accelerate customer onboarding, strengthen fraud detection, reduce operational costs and extend banking services to underserved populations.
However, I believe the discussion surrounding AI in banking often places too much emphasis on technological capability and too little on institutional responsibility.
A bank is not an ordinary technology company. It holds public deposits, manages sensitive financial information, facilitates essential payments and makes decisions that affect the economic opportunities of individuals and businesses.
Errors in these decisions can have serious consequences.
My central assessment is that AI will create lasting value in Indian banking only when automation is accompanied by strong governance, effective security, explainable decisions and meaningful human accountability.
1. Why Artificial Intelligence Matters for Indian Banking
India has developed extensive digital financial infrastructure.
UPI, digital identity verification, account aggregation, mobile banking and electronic payments have expanded the volume and variety of information flowing through the financial system.
Banks increasingly process large quantities of transaction data, customer documents, payment information, credit applications and operational records.
Traditional software can handle structured transactions efficiently.
However, many banking activities also involve interpreting complex information or recognising patterns that are difficult to capture through fixed rules.
AI can support such activities by analysing data at scale.
For example, an AI-assisted system may identify unusual transaction patterns more rapidly than manual monitoring, or help credit officers review numerous financial documents.
The opportunity extends beyond speed.
Properly designed systems may improve consistency, reduce repetitive work and help banks identify risks earlier.
However, technological capability should not automatically be equated with business value.
A system that processes loan applications faster but produces more inaccurate credit decisions does not represent genuine progress.
Similarly, a fraud detection system that blocks large numbers of legitimate payments may reduce certain fraud exposures while creating serious customer difficulties.
In my view, the success of AI should be measured by improvements in accuracy, fairness, reliability, customer outcomes and risk management—not simply by the number of tasks automated.
2. AI-Powered Lending: Faster Decisions, but Are They Better?
Credit assessment is one of the most promising applications of AI.
Traditional lending involves examining income, employment, repayment history, existing obligations, financial statements and other borrower information.
Banks increasingly use statistical models and automated systems to support these processes.
AI may extend their capabilities by recognising more complex relationships between borrower characteristics and repayment outcomes.
Potential Benefits
AI-assisted lending could help banks:
- Analyse loan applications more efficiently.
- Identify inconsistencies in financial documents.
- Detect potentially fraudulent applications.
- Estimate repayment risks.
- Support credit monitoring.
- Reduce repetitive manual processing.
- Serve eligible customers with limited conventional credit histories.
For example, a small business may lack extensive audited financial records but possess legitimate transaction and cash-flow information.
Where the data is lawfully available and appropriate for lending assessment, analytical tools may help lenders understand the business's financial position.
This could support more informed consideration of borrowers who are difficult to assess through traditional methods.
However, there are significant limitations.
The Problem of Incomplete Data
An AI system learns from the information available to it.
If historical lending data disproportionately represents customers who previously qualified for formal credit, the system may struggle to assess applicants with different financial backgrounds.
An apparently accurate model may therefore reproduce historical lending patterns rather than identify genuinely creditworthy borrowers.
The Problem of Automated Rejection
Suppose a loan application is rejected because an algorithm assigns the borrower a high-risk classification.
If the institution cannot identify the main factors influencing that assessment, its ability to investigate potential errors becomes limited.
A model-generated score should therefore not be treated as an unquestionable decision.
My assessment: AI should strengthen the evidence available to credit officers, while banks retain responsibility for lending decisions, applicable disclosures and customer protection.
3. Fraud Detection: One of AI's Strongest Practical Applications
Banks process enormous volumes of financial transactions.
Fraudulent activity can involve account takeovers, unauthorised transfers, impersonation, mule accounts, forged documents and coordinated transaction networks.
Traditional fraud detection systems often rely on predefined rules.
For example, a system may flag transactions above a threshold or payments to certain high-risk destinations.
AI can potentially identify more complicated patterns.
It may examine combinations of transaction frequency, beneficiary relationships, device characteristics, transaction timing and deviations from established behaviour.
Example: Suspicious Transaction Monitoring
Consider a bank account that normally receives a monthly salary and makes routine household payments.
Suddenly, the account begins receiving numerous unrelated transfers and rapidly forwarding them to several beneficiaries.
An AI-assisted monitoring system may recognise this behaviour as unusual and refer it for further investigation.
However, unusual activity is not necessarily fraudulent.
A customer operating a legitimate seasonal business could also experience sudden changes in transaction patterns.
The distinction matters.
False Positives and False Negatives
Fraud detection systems face two important errors.
False positive: A legitimate transaction is incorrectly identified as suspicious.
False negative: A fraudulent transaction is not detected.
Excessively aggressive models may inconvenience genuine customers through unnecessary blocks and repeated verification.
Overly permissive models may fail to prevent substantial losses.
Banks must therefore balance security with customer access.
Fraud detection systems should be evaluated using detection quality, false-positive rates, response time, operational costs and customer impact.
A successful AI fraud prevention programme should reduce actual financial harm without unnecessarily restricting legitimate banking activity.
4. Operational Efficiency: The Less Visible AI Opportunity
AI in banking is often discussed in connection with chatbots and automated lending.
However, I believe some of its most sustainable benefits may emerge in less visible internal processes.
Banks operate through thousands of repetitive administrative activities.
These include:
- Processing customer documents
- Reviewing loan files
- Reconciling accounts
- Classifying complaints
- Examining audit observations
- Analysing regulatory correspondence
- Monitoring operational exceptions
- Preparing internal management reports
Many of these functions involve structured information, repetitive interpretation and document-intensive workflows.
AI-assisted systems may reduce the time required to complete certain activities.
For example, a document-processing system could extract relevant information from a loan application and identify missing documents.
A compliance support system could help staff locate relevant regulatory instructions.
A complaint-analysis system could group similar grievances and identify recurring operational problems.
Efficiency Is More Than Staff Reduction
The business case for automation is frequently framed around reducing employee costs.
I consider this an incomplete perspective.
Banking efficiency should also be assessed through accuracy, service consistency, turnaround time and the capacity of employees to address more complex customer needs.
A bank that automates routine work but fails to improve customer service may have achieved cost savings without meaningful institutional improvement.
The more valuable outcome is to free skilled employees from repetitive activity so they can concentrate on judgment-intensive work.
5. Generative AI: Useful Assistance With Important Limitations
Generative AI introduces additional possibilities.
Unlike many conventional machine-learning systems, generative models can produce natural-language responses, draft documents, summarise information and assist with analytical tasks.
Potential banking applications include:
- Customer-service assistance
- Internal knowledge search
- Regulatory document summarisation
- Software development support
- Employee training
- Operational reporting
- Draft preparation
- Complaint classification and routing
For example, an internal assistant could help a bank employee locate instructions applicable to a particular service request.
However, generative AI presents a serious challenge: it can produce convincing but incorrect information.
A system may invent regulatory requirements, misinterpret a banking circular or provide an inaccurate explanation of a financial product.
Such errors can be particularly dangerous when customers rely on the information to make financial decisions.
Why Human Verification Matters
A generated answer should not be treated as authoritative simply because it is fluent and technically sophisticated.
For sensitive banking uses, organisations should maintain verified information sources, testing procedures and escalation mechanisms.
Important outputs may require human review before being relied upon.
In my view, generative AI is especially valuable as a knowledge and productivity assistant, but high-impact decisions require stronger controls.
The ability to produce an answer is not equivalent to the ability to guarantee its correctness.
6. Model Bias: The Risk of Automating Historical Inequality
Algorithmic bias is one of the most important concerns surrounding AI-based financial decision-making.
A model may produce unequal outcomes because of biased historical data, inappropriate variables, inaccurate assumptions or weaknesses in the development process.
Example: Credit Assessment
Suppose historical lending records contain relatively few successful borrowers from a particular region.
An AI system trained on these records may estimate elevated risk for applicants from that area.
However, the limited historical lending record may reflect past access barriers rather than genuine differences in repayment capacity.
The model could therefore disadvantage potentially creditworthy applicants.
Such problems are not always obvious.
Even when sensitive personal characteristics are excluded, other variables may act as indirect proxies.
What Responsible Banks Should Do
Institutions should examine:
- Data representativeness
- Model performance across relevant customer groups
- Unintended discrimination
- Proxy variables
- Model assumptions
- Changes in borrower populations
- Error rates across segments
- Mechanisms for reviewing disputed outcomes
Fairness testing should be proportionate to the model's purpose, risk and applicable law.
My position is that financial inclusion should not be judged by whether AI can approve more loans, but by whether it can support fairer, more accurate and sustainable credit access.
7. Data Privacy: Financial Information Requires Exceptional Care
AI systems depend heavily on data.
In banking, this information may include financial transactions, account balances, identity records, repayment behaviour and other personal details.
The ability to process this information creates valuable opportunities, but also substantial privacy risks.
A system may collect more information than necessary, combine datasets in unexpected ways or expose sensitive information through insecure integrations.
For example, a bank should not assume that every piece of customer information it possesses is automatically appropriate for training an AI model.
The purpose, lawful basis, customer expectations and applicable regulatory restrictions matter.
India's Regulatory Environment
The Digital Personal Data Protection Act, 2023 and the Digital Personal Data Protection Rules, 2025 establish an important framework for personal-data governance in India, with provisions becoming effective according to their notified commencement arrangements.
Banks must also consider RBI requirements applicable to their activities.
For digital lending, RBI's 2025 Directions address data collection, consent, use, sharing, storage and customer disclosures.
These requirements are particularly relevant where banks work with digital lending platforms and lending service providers.
Privacy by Design
I believe banks should apply the principle of collecting and processing only the information necessary for a legitimate, defined purpose.
Responsible AI programmes should consider:
- Data minimisation
- Clear processing purposes
- Access controls
- Appropriate consent or other lawful grounds
- Data retention and deletion
- Secure handling of customer information
- Vendor access restrictions
- Audit trails
- Incident response
Privacy should be built into AI architecture rather than treated as a documentation requirement after deployment.
8. AI Creates New Cybersecurity Threats
AI can strengthen cybersecurity, but it can also introduce new attack surfaces.
Banks should distinguish between using AI to defend systems and securing the AI systems themselves.
Data Poisoning
Attackers may attempt to manipulate training or operational data so that a model produces unreliable results.
In fraud detection, contaminated data could weaken the system's ability to identify suspicious activity.
Adversarial Manipulation
Attackers may craft inputs designed to exploit weaknesses in AI models.
A transaction monitoring system could potentially be manipulated if its behaviour becomes sufficiently predictable.
Prompt Injection
Generative AI applications may process documents, emails or other external information containing malicious instructions.
If a system improperly treats such content as authoritative, it could disclose information or perform unintended actions.
Sensitive Data Leakage
AI assistants may expose confidential customer information through insecure retrieval, logging, integrations or output handling.
Deepfake and Identity Fraud
AI-generated voices, images and videos can make impersonation attempts more convincing.
Banks may need stronger verification and fraud response mechanisms as these technologies advance.
Model Supply Chain Risk
An organisation may depend on externally developed models, software libraries, cloud platforms and specialised service providers.
Weaknesses in these components can affect banking operations.
The critical lesson is that AI security cannot be separated from broader cybersecurity, technology outsourcing and operational resilience.
9. Third-Party AI and Cloud Dependence: Who Is Accountable?
Many banks will not build every AI system internally.
Instead, they may procure software, use external models or integrate specialised services provided by technology vendors.
This approach can accelerate implementation and reduce development costs.
However, outsourcing creates dependencies.
A bank may rely on a service provider for model hosting, data processing, document analysis or customer interaction.
If the provider experiences a cyber incident, service interruption or material model failure, the bank's operations may be affected.
Vendor Risk Questions
Before relying on an external AI service, banks should examine:
- Who controls and accesses customer data?
- Where is data processed and stored?
- Can the model's decisions be explained?
- Is independent testing possible?
- How are incidents reported?
- What happens if the provider becomes unavailable?
- Can the bank migrate or discontinue the service?
- What operational fallback arrangements exist?
Contractual arrangements should support auditability, security, business continuity and appropriate regulatory oversight.
In my assessment, outsourcing technology does not outsource the bank's responsibility for customer outcomes or risk management.
10. Human Oversight Must Be Meaningful, Not Symbolic
Human oversight is frequently presented as a safeguard against AI-related risks.
However, simply placing a person somewhere in an automated workflow does not guarantee effective supervision.
Consider a system that rejects thousands of loan applications automatically.
If an employee reviews only the final result without access to the relevant information, model reasoning or authority to change the decision, the review may have little practical value.
Meaningful oversight requires more.
Human-in-the-Loop
A designated employee reviews or approves a consequential action before it is finalised.
Human-on-the-Loop
Employees supervise an automated process and intervene when specified conditions arise.
Human-in-Command
The institution retains authority to modify, suspend or discontinue automated systems.
The appropriate approach depends on the risk and significance of the task.
High-volume low-risk administrative activity may need different controls from automated lending, fraud-related account restrictions or customer grievance decisions.
The Automation Bias Problem
Employees may become overly confident in algorithmic recommendations.
If an AI system usually performs well, people may stop questioning its outputs.
This is known as automation bias.
Banks must therefore train employees to recognise model limitations, investigate unusual results and exercise independent judgment.
A human reviewer is valuable only when that person understands the decision, has sufficient information and possesses genuine authority to intervene.
11. RBI's FREE-AI Framework: A Foundation for Responsible Adoption
In August 2025, the RBI published the report of its Committee on the Framework for Responsible and Ethical Enablement of Artificial Intelligence in the Financial Sector, commonly called FREE-AI.
The report aims to balance the benefits of AI with the need to manage emerging risks.
It identifies seven guiding principles, described as Sutras:
- Trust Is the Foundation
- People First
- Innovation over Restraint
- Accountability
- Fairness and Equity
- Understandable by Design
- Safety, Resilience and Sustainability
The report also presents recommendations covering innovation infrastructure, policy, capabilities, governance, protection and assurance.
These principles reflect an important regulatory perspective.
AI innovation and risk management should not be treated as opposing objectives.
Responsible governance can make technological adoption more sustainable.
Regulatory Developments in 2026
In June 2026, the RBI proposed model risk management guidelines covering AI and machine-learning systems.
The proposed approach included stronger model governance, independent validation, model inventories, risk monitoring and human oversight.
RBI communications in August and September 2026 further emphasised accountability, fairness, cybersecurity and the ability to intervene when automated systems behave unexpectedly.
An important distinction must be maintained: committee recommendations and draft guidelines are not automatically equivalent to enforceable final regulatory obligations.
Banks should verify the current status of specific directions before treating proposed measures as mandatory.
12. The Model Risk Management Challenge
Every AI model has limitations.
Its performance depends on its training data, design, assumptions and operating environment.
A model that performs well during stable economic conditions may behave differently during a recession.
A fraud model trained on historical patterns may become less effective as criminal behaviour evolves.
A customer-service model may struggle with new products or recently amended regulations.
Model Drift
Model drift occurs when changes in data or underlying relationships reduce a model's reliability over time.
For example, a lending model developed using historical repayment patterns may become less accurate after a significant change in interest rates or employment conditions.
Model Validation
Banks should test models before deployment and review them periodically.
Validation may examine:
- Data quality
- Accuracy
- Stability
- Fairness
- Explainability
- Security
- Performance under adverse conditions
- Suitability for the intended purpose
Model Inventory
An institution should know which AI systems it uses, what decisions they support, who owns them and what risks they introduce.
Without an accurate model inventory, management may be unable to assess cumulative dependence on automated systems.
Model Retirement
AI systems should not operate indefinitely simply because they were successfully deployed.
Banks need procedures for restricting, updating, replacing or discontinuing models that become unreliable.
In my view, model risk management should be treated as an ongoing operational discipline rather than a one-time technology approval.
13. Explainability and Customer Rights
A major challenge in AI-enabled banking is the ability to explain decisions.
Customers may want to know why a loan was rejected, a transaction was blocked or a risk classification was applied.
Banks cannot always provide detailed technical explanations of proprietary models.
However, this does not eliminate the need for meaningful communication.
Where legally required or operationally appropriate, institutions should provide understandable information about relevant decision factors and available review processes.
For example, if a lending decision is based on insufficient documented income or excessive existing debt obligations, the customer should not necessarily be confronted with an unexplained algorithmic label.
Explainability is also important internally.
Risk managers, auditors and senior management need enough information to understand whether automated systems are operating as intended.
A model that cannot be adequately evaluated is difficult to govern.
14. AI and Banking Employment: Transformation Rather Than a Simple Replacement Story
AI will influence banking roles.
Routine data entry, document classification, basic enquiry handling and repetitive reconciliation activities may become increasingly automated.
However, the broader implications for employment are more complex.
Banks will continue to require professionals capable of interpreting financial information, assessing risk, handling customers, investigating fraud, managing technology and making accountable decisions.
New skills will become more valuable.
These include:
- Data literacy
- Model risk management
- Cybersecurity
- AI governance
- Business process design
- Digital customer support
- Technology audit
- Regulatory technology
- Human-machine workflow management
Institutions should invest in retraining and skill development rather than viewing technology solely as a mechanism for workforce reduction.
I believe the most effective banks will redesign work so that technology handles appropriate repetitive tasks while employees focus on judgment, complex problem-solving and customer relationships.
15. AI Adoption Will Not Be Equal Across Indian Banks
Large commercial banks generally have greater resources to invest in technology, specialised employees and infrastructure.
Smaller banks and financial institutions may face higher relative costs.
They may also depend more heavily on external vendors.
This creates the possibility of unequal AI adoption across the financial sector.
Large Banks
Potential advantages include extensive datasets, established technology infrastructure and specialist teams.
However, large institutions also operate complex systems and face significant integration challenges.
Smaller Banks
Smaller institutions may benefit from affordable cloud-based solutions and shared technology capabilities.
However, vendor dependence, limited internal expertise and governance capacity require attention.
Cooperative and Regional Banking
AI tools may support operational efficiency and service delivery, but they must be adapted to the needs and capabilities of the institutions concerned.
A complex AI system is not automatically appropriate for every bank.
My assessment is that responsible adoption should be proportionate to institutional size, technical capacity, risk exposure and the actual business problem being addressed.
16. How Should Banks Measure AI Success?
I would discourage banks from reporting AI progress solely through the number of deployed models, chatbot interactions or automation projects.
A stronger measurement framework should assess whether AI improves banking outcomes.
| Area | Useful Performance Indicators |
|---|---|
| Lending | Processing time, default performance, fairness and credit-decision quality |
| Fraud detection | Actual fraud losses, detection rates, false positives |
| Operations | Error rates, turnaround time, cost per transaction |
| Customer service | Resolution quality, complaint recurrence, customer satisfaction |
| Security | AI-related incidents, vulnerabilities, recovery time |
| Model governance | Validation status, drift alerts, unresolved model issues |
| Financial inclusion | Responsible access, customer outcomes and repayment sustainability |
These measures should be assessed relative to suitable baselines.
A bank should be able to explain what improved after introducing AI and what additional risks or costs emerged.
If a project does not produce measurable benefits, its continued operation should be reconsidered.
17. Seven Strategic Priorities for Responsible AI in Indian Banking
In my assessment, the following priorities deserve particular attention.
Priority 1: Establish Board-Level AI Governance
Banks should define responsibility for AI strategy, risk appetite, model oversight and incident escalation.
Priority 2: Apply Risk-Based Controls
Not every AI application requires identical controls.
High-impact use cases such as credit decisions, transaction restrictions and fraud investigations require stronger validation and oversight.
Priority 3: Improve Data Quality and Privacy
Reliable models require reliable data.
Banks should minimise unnecessary data collection and protect customer information throughout its lifecycle.
Priority 4: Strengthen Independent Validation
Models should be tested by appropriately qualified teams with sufficient independence from their development and deployment.
Priority 5: Maintain Human Intervention Capabilities
Banks must be able to review, override, suspend and replace automated decisions where appropriate.
Priority 6: Secure the AI Supply Chain
Third-party models, cloud services, external APIs and technology vendors require comprehensive risk evaluation.
Priority 7: Invest in Employee Capability
Responsible AI cannot be achieved through software procurement alone.
Employees, risk managers, auditors and senior leadership need sufficient understanding of AI capabilities and limitations.
18. Three Possible Futures for AI in Indian Banking
The long-term impact of AI will depend heavily on the choices institutions make today.
Scenario 1: Responsible Transformation
Banks integrate AI with strong governance, reliable data, independent validation and effective human oversight.
Efficiency improves, fraud monitoring strengthens and customers receive better services.
AI becomes an enabling capability rather than an uncontrolled source of risk.
Scenario 2: Rapid Automation With Weak Controls
Institutions prioritise deployment speed and cost reduction.
Models are insufficiently tested, employees become overly dependent on automated outputs and customer disputes become difficult to resolve.
Operational and reputational risks increase.
Scenario 3: Fragmented Adoption
Large institutions develop advanced AI capabilities, while smaller banks face constraints involving expertise, investment and technology dependence.
Uneven adoption creates differences in operational efficiency and risk management capacity.
These are analytical scenarios, not predictions.
Their purpose is to illustrate how governance and institutional investment may shape outcomes.
19. My Outlook: Banking Needs Responsible Intelligence, Not Just Artificial Intelligence
I believe artificial intelligence will become a significant component of Indian banking infrastructure.
Its applications in fraud detection, lending, operations, customer service and compliance are too important to ignore.
However, I do not consider technological sophistication a sufficient measure of institutional progress.
A bank can deploy advanced AI models and still provide poor customer service.
It can automate credit decisions and still misjudge borrower risk.
It can introduce sophisticated fraud detection while remaining vulnerable to technology failures and cyberattacks.
The fundamental challenge is therefore organisational rather than purely technical.
Banks must establish governance structures that enable them to understand, test and control the systems they deploy.
They must also invest in employees capable of recognising when automated outputs are inappropriate.
The objective should be to combine machine-assisted analytical capability with sound banking judgment.
In my assessment, the future competitive advantage will belong not to institutions that automate the greatest number of decisions, but to those that make better, safer and more accountable decisions through responsible use of technology.
Conclusion: Artificial Intelligence Must Strengthen Banking Trust
AI represents a substantial opportunity for India's banking industry.
It can help improve productivity, analyse financial information, identify fraudulent behaviour, support credit decisions and make banking services more accessible.
Yet these benefits cannot be separated from the risks of algorithmic bias, data misuse, cyber vulnerability, opaque decision-making and excessive automation.
Banking institutions have obligations that extend beyond technological performance.
They are responsible for customer funds, sensitive information, fair treatment and confidence in the financial system.
For this reason, AI governance must become an integral part of banking governance.
Cybersecurity, privacy, model validation and human oversight should be embedded into technology design and operational processes from the beginning.
My central conclusion is simple: The future of AI in Indian banking should not be defined by how much human judgment can be removed, but by how effectively artificial intelligence can strengthen human judgment, institutional accountability and customer trust.
Author's perspective: This Expert Insight presents a technology-focused assessment of AI adoption in Indian banking. The analysis and forward-looking judgments are editorial opinions informed by publicly available regulatory material and technology risk-management principles.
References
RBI — FREE-AI Committee Report, August 2025. Detailed recommendations on AI governance, fairness, explainability, data quality and model risk.
Reserve Bank of India
RBI — A Vision for Responsible AI, Resilient Banking, August 2026. Discusses human judgment, governance, innovation and operational resilience.
Reserve Bank of India
RBI — AI Risk, Cybersecurity and Human Accountability. Examines adversarial risks, model governance and meaningful human oversight.
Reserve Bank of India
RBI — Digital Lending Directions, 2025. Regulatory requirements concerning digital lending, borrower assessment, data collection, consent and customer protection.
Reserve Bank of India
Government of India — DPDP Rules, 2025. India's personal data protection framework and implementation provisions.
Press Information Bureau
RBI — Commercial Banking Master Directions. Current regulatory directions concerning banking cybersecurity, technology risk and digital payment security.
Master Directions



