RBI embraces AI to tackle fraud and enhance efficiency
Spread across the Reserve Bank of India’s latest annual report published on May 29, between updates on inflation, liquidity, and financial stability, was a two-letter word that kept surfacing - AI.
Appearing close to 20 times in the document, AI wasn’t just a buzzword peppered in for effect. It was everywhere — embedded in complaints redressal systems, powering fraud detection models, guiding supervisory frameworks, in RBI’s international policy submissions, and even shaping internal workflows through the central bank’s own generative AI tool.
Enter ChiRAG: The brain behind the curtain
ChiRAG, short for Chat Interface with Retrieval Augmented Generation — is the Reserve Bank’s very own generative AI platform. “The potential of emerging technologies, particularly generative AI which can generate context-aware, human-like responses and analyse vast amounts of data, is rapidly gaining traction in the central banking landscape,” the report notes.
ChiRAG, it adds, was initially designed as “a tool for information extraction and synthesis,” but is now being developed further into “a sophisticated orchestration layer, which will seamlessly coordinate with diverse types of information and data associated with the Reserve Bank’s wide array of functions.”
A smarter way to complain
If ChiRAG hints at a shift in how the RBI processes information internally, another effort points to how it wants to engage externally. The Reserve Bank is “collaborating closely with Reserve Bank Information Technology Pvt. Ltd. (ReBIT) to integrate artificial intelligence (AI) into the complaint management system in a phased manner.” In Phase I, a conversational AI chatbot will assist complainants, while Phase II will involve “more advanced features for processing the complaints.”
The initiative suggests a regulator thinking deeply about user experience — and reworking redress systems to be faster, smarter, and scalable.
Stopping fraud before it starts
Fraud detection, particularly in the digital payments space, emerged as another priority. “To protect customers from digital payment frauds,” the report notes, “the Reserve Bank constituted a committee to examine various aspects of setting up a Digital Payments Intelligence Platform (DPIP) to harness advanced technologies for the purpose.”
Reserve Bank Innovation Hub (RBIH) has been tasked with building a prototype of the DPIP in consultation with five to ten banks. While the report does not detail the exact role of AI within this platform, the term “advanced technologies” in this context sits comfortably within the bank’s broader AI-led approach to fraud surveillance.
This is not the only project underway. RBIH has also developed a specialised supervised machine learning tool — MuleHunter.ai — to tackle a problem plaguing the digital ecosystem: mule bank accounts. “The model leverages advanced AI/ML techniques to learn patterns of mule account activity from data, achieving higher accuracy as compared to the traditional systems,” the report states. The tool is being tested and deployed in a few large public sector banks, and is aimed at enabling “near-real-time identification of mule accounts.”
Making banks accountable — with data
The Reserve Bank is also building an ecosystem of structured assessments. The Consumer Protection Assessment Matrix (CoPAM) — now being piloted — evaluates regulated entities on metrics like complaint handling time and repeat issues.
Meanwhile, tools like the Supervisory Data Quality Index (sDQI) are enhancing data completeness and consistency. While neither is explicitly marked as AI-led, both reflect the regulator’s shift toward metric-driven supervision that may well support AI integration down the line.
Watching the system, smarter
Beyond customer-facing tools, the RBI is laying the foundation for deeper AI-driven supervision. The report confirms that a dedicated Advanced Supervisory Analytics Group (ASAG) has been set up by the Department of Supervision “for increasing the use of techniques like AI/ML in the supervisory process.”
The group has already developed a suite of analytics models, including microdata analytics, governance assessment, social media monitoring, a fraud vulnerability index, borrowers’ vulnerability model, and an asset quality prediction model. More such models are under development.
These tools aren't just experiments; they mark a shift in how the RBI sees the future of systemic oversight — one powered as much by data as by prudence.
Ethics, adoption and audit: AI governance begins
Perhaps the most telling development, though, is RBI’s attention to the ethical side of AI. In December 2024, the central bank constituted an external expert committee to create a Framework for Responsible and Ethical Enablement of AI (FREE-AI). This framework, due for formulation in 2025–26, aims to guide the responsible deployment of AI in the financial sector. In the same vein, the RBI will also explore the feasibility of adopting AI and ML in internal audit processes, and publish a research paper on leveraging AI to enhance legal functions in the financial sector.
Taken together, these developments show how deeply the institution is embedding AI not just as a set of innovations, but as part of its core governance and compliance strategy.
Mapping the ecosystem, globally and locally
AI isn’t being integrated only by the RBI — it’s also being deployed across the industry. On May 28, 2024, the central bank launched two key repositories: the FinTech Repository, which captures tech stack information from fintech entities, and the EmTech Repository, focused on how regulated entities are using AI, ML, cloud, and DLT.
Both are managed by RBIH, and are designed to help “track sectoral trends and generate analytics that will be useful for both policymakers and the industry.”
And on the international stage, the Department of Economic and Policy Research contributed to global regulatory dialogues under the Financial Stability Board (FSB), including inputs on AI and tokenisation as part of its engagement with cross-border financial innovation and resilience frameworks.
Not a race — a redesign
The tone of the report is deliberate discussing through terms like “phased manner,” “pilot testing,” and “collaborative frameworks.” The goal isn’t speed. It’s structural integration.
This year’s annual report may not declare an AI revolution. But it marks something quieter and more consequential: an institutional shift. From regulatory tools like ChiRAG and MuleHunter to frameworks like FREE-AI and fraud detection platforms like DPIP, RBI’s AI playbook is no longer emerging — it’s being implemented.If AI was once on the regulator’s radar, it’s now on its roadmap.