Reserve Bank of India (RBI) Governor Sanjay Malhotra has urged Indian banks to embrace artificial intelligence (AI) as a significant instrument to rework lending, monetary inclusion, customer support, operational effectivity and fraud detection. He additionally warned that fast AI adoption should be backed by robust governance, explainability, information privateness, cybersecurity and human oversight.
Whereas talking on the FIBAC 2026 convention in Mumbai, RBI Governor Malhotra stated AI might outline the 2020s for Indian banking in the identical means liberalisation formed the Nineteen Nineties and digitalisation remodeled the 2010s.
“Synthetic Intelligence isn’t a single expertise to be procured, nor a venture to be accomplished. It’s a new means of doing enterprise, of operating a financial institution,” Malhotra stated.
He stated banks to resolve whether or not they’ll intentionally form their AI journey or permit the expertise to form them by default.
AI can develop entry to credit score
Malhotra stated AI might basically change the economics of credit score supply, notably for debtors who’ve restricted formal monetary histories.
Conventional lending fashions can wrestle to evaluate new-to-credit debtors, gig staff and small companies with out formal books. AI fashions can as a substitute use various info akin to money flows, GST filings, utility funds and digital footprints.
This might considerably develop the pool of debtors thought-about “bankable” by lenders whereas reducing the marginal value of underwriting loans.
AI-powered credit score threat fashions might additionally assist banks establish rising stress earlier. Liquidity forecasting and situation evaluation might permit lenders and regulators to detect issues earlier than they seem in conventional monetary statements.
AI might increase monetary inclusion
The RBI Governor stated AI might change into a significant accelerator of economic inclusion in India. Voice-based interfaces in Indian languages might make banking simpler for purchasers who face language limitations. Predictive fashions might additionally establish debtors vulnerable to default early sufficient for banks to supply help as a substitute of merely starting restoration motion.
India’s current digital public infrastructure provides the nation a bonus in deploying AI, Malhotra stated.
He pointed to Aadhaar, UPI, DigiLocker, ONDC, Account Aggregator and the Unified Lending Interface (ULI) as platforms that may help additional innovation.
“AI, layered on prime of this stack, has the potential to do for monetary judgment what UPI did for monetary transactions: make it instantaneous, granular, and obtainable to the final mile,” he stated.
AI can enhance customer support and cut back prices
AI also can assist banks enhance customer support and productiveness, in accordance with Malhotra.
AI-assisted relationship managers might establish appropriate merchandise and threat alerts, permitting workers to serve extra prospects. AI might additionally help grievance redressal and supply personalised monetary steering.
Banks might use AI to automate doc processing, reconciliation and inner audit sampling. The expertise might additionally automate transaction reporting and regulatory returns.
This might decrease compliance and operational prices whereas permitting expert workers to give attention to work that requires human judgment.
AI may also help banks combat fraud
Malhotra additionally highlighted AI’s function in combating digital fraud. “Fraud right now strikes on the velocity of an API name,” he stated, including that conventional rules-based fraud methods can wrestle to maintain tempo with criminals who continually change their strategies.
Machine-learning fashions can repeatedly study from transaction patterns and establish anomalies in actual time.
This might permit banks to make use of AI to detect fraudulent transactions earlier than losses materialise.
RBI flags seven key dangers from AI
Whereas the RBI helps better AI adoption, Malhotra warned that banks should handle the dangers related to the expertise.
The primary is the “black field” downside. Superior AI and generative AI fashions might not at all times clarify how they reached a specific choice.
If an AI system rejects a mortgage software, as an illustration, each the client and regulator ought to have the ability to perceive the rationale behind the choice.
The second threat is bias and exclusion. AI methods educated on historic lending information might reproduce current biases towards sure geographies, occupations or communities.
Malhotra stated equity should be constructed into AI methods from the start quite than handled as a compliance requirement.
The RBI additionally warned about focus and herding dangers. If a number of banks depend on the identical AI fashions or expertise distributors, a typical error or vulnerability might change into a system-wide downside.
Related AI-based buying and selling fashions might additionally trigger banks to behave in the identical means during times of market stress, probably rising volatility.
Banks should handle vendor and information dangers
Third-party dependence is one other concern. Smaller banks might depend on expertise distributors quite than develop AI fashions themselves.
Malhotra stated outsourcing preparations should embrace AI-specific accountability, audit rights, the flexibility to hunt explanations and a reputable exit plan if a vendor or mannequin must be changed.
Information privateness and safety will even change into more and more necessary as AI methods require massive quantities of knowledge.
“The Digital Private Information Safety Act is the ground, not the ceiling, of what prospects ought to count on from their financial institution,” Malhotra stated.
The RBI additionally flagged cyber dangers, together with information poisoning, mannequin manipulation and assaults designed to idiot AI-based fraud detection methods.
‘The mannequin determined’ can’t be an excuse
The RBI Governor’s strongest warning involved the attainable erosion of human judgment and accountability. He stated banks can’t shift duty for choices to algorithms, no matter how subtle the expertise turns into.
“The mannequin determined’ can by no means be a suitable reply to a buyer, an auditor, or the Reserve Financial institution,” Malhotra stated.
Banks should retain the flexibility to elucidate AI choices, intervene when vital and override a mannequin the place required.
RBI asks banks to strengthen AI governance
The RBI needs banks to deal with AI governance as an instantaneous precedence quite than a future compliance train.
Banks ought to keep an entire stock of each AI system they use, together with AI embedded in merchandise equipped by distributors. They need to additionally set up board-approved AI governance insurance policies with clear accountability for outcomes.
The central financial institution needs lenders to construct the flexibility to elucidate AI-driven choices that materially have an effect on prospects, notably in lending and fraud.
Banks also needs to red-team and stress-test AI methods earlier than deployment and periodically afterwards.
Human oversight should stay in place wherever an AI error might trigger materials hurt to prospects or monetary stability.
RBI to comply with a principles-based strategy
Malhotra stated the RBI will take a principles-based and proportionate strategy to regulating AI quite than impose inflexible guidelines.
The dangers confronted by a big financial institution utilizing proprietary AI fashions might differ considerably from these confronted by a smaller lender utilizing an off-the-shelf product from a expertise vendor.
The RBI additionally plans to work with banks as AI expertise evolves as a substitute of regulating it from a distance.
The central financial institution will proceed to make use of its regulatory sandbox to supply a protected atmosphere for testing modern use instances.
It additionally plans to facilitate frequent utilities akin to MuleHunter and the proposed Digital Funds Intelligence Platform to strengthen fraud detection.
RBI stays centered on monetary stability and credit score development
Malhotra additionally outlined progress on the RBI’s broader regulatory priorities introduced at FIBAC 2025.
The central financial institution has taken measures masking credit score threat capital, the anticipated credit score loss framework, venture finance, related-party transactions, dividend coverage and internet open positions.
The RBI stays on monitor to implement relevant Basel III pointers from April 1, 2027, on a calibrated glide path.
On ease of doing enterprise, the central financial institution has diminished regulatory burdens on boards, consolidated regulatory directions, streamlined types of enterprise and rationalised current-account and working-capital norms.
It has additionally continued to strengthen digital public infrastructure akin to Account Aggregator and ULI to decrease the price of credit score origination and underwriting, notably for MSMEs and underserved debtors.
Banks that perceive AI can have an edge
Malhotra stated India has vital alternatives to make use of AI to enhance credit score entry, customer support, monetary inclusion and fraud detection.
However the RBI doesn’t need banks to pursue AI adoption merely for the sake of adopting the expertise.
“The banks that can win within the AI period won’t essentially be those that undertake essentially the most AI, or the quickest,” Malhotra stated.
As an alternative, he stated banks that perceive what they’re deploying, keep clear accountability and shield buyer belief will probably be higher positioned to achieve the AI period.
