RBI Launches AI-Powered Intelligence Platform to Combat Digital Payment Fraud

According to Storyboard18, citing a CNBC TV18 report, the Reserve Bank of India is preparing to give banks access to a Digital Payment Intelligence Platform to strengthen the financial system’s response to increasingly sophisticated digital payment fraud. RBI Governor Sanjay Malhotra announced the initiative at FIBAC 2026 in Mumbai, and said artificial intelligence was becoming relevant on both sides of the fraud equation. For digital banks, neobanks and BaaS providers, the event is less a product launch than a change in the control architecture: the institution remains accountable for what its automated systems do.

A fraud platform with an accountability condition

Malhotra described a two-sided problem. Financial institutions are expanding their use of AI across fraud monitoring, lending and customer services, while fraudsters are using similar technologies. His starting point was not a promise of better detection. Banks were first expected to understand how widely AI was already deployed and then establish appropriate safeguards around those systems.

The governor called for a detailed record of the AI systems used across bank operations. He also recommended that boards formally approve AI governance frameworks and that models be tested through red-team exercises and stress scenarios. The sequence is significant: inventory, oversight and testing come before confidence in automated output. Without those steps, a model is not a control framework; it is another dependency whose failure path has not been made institutionally visible.

Malhotra was explicit about liability. Responsibility for banking decisions cannot be transferred to an AI system, he said. Banks must remain accountable, with human intervention continuing to play a role where appropriate. For a BaaS provider, the implication is straightforward: model ownership cannot be outsourced with model hosting, and the regulated institution’s exposure does not end at the vendor interface.

Smaller lenders and the vendor boundary

The governor singled out smaller lenders. They may not have the resources to build their own AI models and may instead depend on external technology providers. In that setting, he said, banks need stronger controls over how customer information is handled and how third-party systems are governed.

That is the practical risk in the announcement. Vendor access does not remove the bank’s responsibility for a system it uses. The relevant questions extend beyond model performance: what customer information is handled, how the system is governed, what records are kept, how the model is tested, and where human intervention continues to play a role. The announcement does not indicate that dependence on an external provider reduces the bank’s accountability.

For neobanks and embedded-finance businesses, this is where the architecture becomes commercially material. Outsourcing model development may reduce the need to build an in-house system, but the governance burden does not fall at the same rate. The RBI’s direction places responsibility on the bank even where the model is supplied by someone else. Due diligence must therefore cover the operating chain, not merely the fraud-detection score produced at the end of it.

The inclusion trade-off

Malhotra also placed the platform within India’s wider digital-finance infrastructure. He pointed to the existing digital financial infrastructure as a foundation for wider AI adoption and said the RBI was working to expand the Unified Lending Interface and Account Aggregator ecosystem. The expansion could provide additional data and infrastructure for technology-led financial services.

The stated benefit is inclusion. Malhotra said AI has the potential to accelerate financial inclusion, but warned that poorly implemented or inadequately monitored automated processes could create fresh barriers for consumers who were already underserved. Automation therefore does not cancel the obligation to design and supervise the system. It changes where that obligation must be exercised.

The announcement was made against a broader regulatory shift. Malhotra cited changes introduced over the past year covering customer service, business operations and the cost of financial intermediation. He also said the RBI had sought to reduce regulatory friction while giving management more flexibility over operational matters; the PRAVAAH regulatory platform had been expanded, including greater automation of application-related processes.

The unresolved issue is implementation. The available report does not provide a timetable or technical specification for access to the new platform, data handling, testing, or the allocation of responsibility. Until those details emerge, the platform should be read as a prospective control dependency rather than a completed fraud solution. The hidden liability is clear: the platform may improve detection, but it cannot repair an institution that has not defined its own records, governance and accountability chain.