Indian banks accelerate AI deployment amid rising cyber risks and regulatory tightening

Indian financial institutions are increasingly integrating AI into core operations amid significant productivity gains and evolving regulatory frameworks, but are also grappling with escalating cyber threats that could reshape the future of digital banking.

Indian banks and non-bank lenders are reaching a point where artificial intelligence is no longer a side experiment in customer service or fraud alerts. It is becoming critical infrastructure, and the latest warning from the Reserve Bank of India shows why that matters. In June 2026, a survey cited in the central bank’s Financial Stability Report found that banks and NBFCs saw AI-enabled cyber threats as the biggest risk to their businesses over the coming 12 months, while preparedness was still uneven and, in many cases, only partly formalised. (livemint.com)

That caution comes after a rapid build-out in adoption. EY said in March 2025 that 74% of Indian financial firms had already started generative AI proof-of-concept work, but only 11% had moved into production-level deployment, a reminder that enthusiasm and scale are not the same thing. The consultancy’s study, based on more than 700 roles across banking and insurance, estimated that generative AI could lift productivity across Indian financial services by 34% to 38% by 2030, with banking operations alone seeing gains of up to 46%. EY also said 42% of organisations were already setting aside dedicated budgets for AI, and that NBFCs and insurers were leading early adoption. (ey.com)

The gains EY expects are not confined to back-office automation. It projects productivity improvements of 38% to 40% in sales and customer service and 34% to 36% in credit and collections, driven by tools such as voice bots, email automation, business intelligence and workflow software. That helps explain why lenders are pushing AI deeper into loan processing, servicing and recovery rather than treating it as a standalone technology project. The big prize is speed at scale: fewer manual hand-offs, faster decisions and more consistent operations across large branch and digital networks. (ey.com)

Credit is where that shift may be felt most sharply. Mint reported in February 2025, citing EY and the Digital Lenders Association of India, that digital lending disbursements stood at about ₹21.6 trillion in FY22 and were expected to reach ₹47.4 trillion by FY26. At an industry event in Mumbai on 6 February 2025, Sunil Mehta, chair of IndusInd Bank, said fintech firms had forced banks to rethink old procurement habits: “That was a game changer.” D. Venkatesh of Lentra said his company’s ambition was to gather permissioned outside-world data and feed it into a rules engine, while Ramesh Narayanaswamy of Aditya Birla Capital argued that under the DPDP Act, “customer consent would be the key”. (livemint.com)

Large lenders have begun to put names and numbers to their programmes. HDFC Bank, according to Mint, has identified more than 15 generative AI initiatives and moved from scattered pilots to what chief executive Sashidhar Jagdishan called a “platform-driven GenAI strategy” aimed at staff productivity and customer service. Jagdishan said the bank had also launched a GenAI Academy because “As GenAI becomes central to both current and future organisational capability, we are investing in developing skills in this space through a structured academy model, offering robust curriculum and tiered proficiency levels”. The same bank has upgraded its cyber security operations centre with AI and machine-learning tools as it confronts phishing, deepfakes and impersonation scams. (livemint.com)

Regulation is starting to catch up with that ambition. The Reserve Bank first set up its committee on the responsible and ethical use of AI in December 2024, then published the FREE-AI committee’s report on 13 August 2025. The report set out seven “Sutras” for adoption and 26 recommendations, including a board-approved AI policy for regulated entities, AI-specific audit and consumer-protection controls, and a requirement that customers should be made aware when they are dealing with AI. It also proposed an AI Innovation Sandbox and a shared financial-sector data infrastructure, signalling that the regulator wants to encourage experimentation, but inside a much tighter governance framework. The report goes further still by classing uses such as credit underwriting and autonomous systems that handle customer interactions, make financial decisions or move customer funds as high-risk. (m.rbi.org.in)

The Reserve Bank’s own projects show the same split between opportunity and control. In its 2024-25 annual report, the central bank said Reserve Bank Innovation Hub had developed MuleHunter.ai, a supervised machine-learning model for near-real-time identification of mule accounts, and that it was being tested and deployed in a few large public sector banks. That is a concrete example of AI being used as a fraud-fighting tool inside the system rather than simply a front-end convenience for customers. At the same time, McKinsey argues that the biggest returns will not come from individual tools alone, but from changing how banks are run. Its March 2025 analysis said banks that simplify processes and apply technology, automation, analytics and training at scale can deliver lasting productivity gains of up to 15% in two years, lifting return on equity by 1.0 to 1.5 percentage points. It pointed to uses ranging from call-centre work and credit memo writing to fraud and dispute handling, research, analytics and software development. (rbi.org.in)

Taken together, the picture is less about a sudden wave of machine-led banking than about a contest over who can industrialise AI safely. Indian lenders have moved well beyond basic chatbots, and the strongest examples now link underwriting, collections, cyber defence, staff training and operating-model redesign. But the gap between pilots and production remains wide, and the RBI’s June 2026 survey suggests many institutions are still building the controls they need. The next winners are unlikely to be those that adopt AI fastest in appearance, but those that can make it reliable, auditable and resilient under regulatory scrutiny. (ey.com)