The Reserve Bank of India (RBI) has released the “Framework for AI Responsibility and Ethical Realization (FreeAI), using a regulatory roadmap containing 26 recommendations, in its most comprehensive policy blueprint for using artificial intelligence (AI) in finance.
This document aims to harness the potential for AI transformation while protecting trust, ethics and stability in the Indian financial system.
Free-AI was founded by RBI in December 2024 and created by a committee chaired by IIT Bombay Professor Pushpak Bhattacharyya.
It investigated global regulations, conducted an industry survey covering more than 600 supervised entities, and, according to an official release, engaged with multiple stakeholders before finalizing the principles.
A study conducted by the committee showed that AI adoption in the financial sector is on a low but increasing. Only 20.8% of the 612 entities surveyed by the committee used or developed AI, and adoptions were biased towards major banks.
Most deployments were in low-risk areas such as customer support, marketing, credit underwriting and cybersecurity tools. Small urban cooperative banks and NBFCs cited costs, data quality and talent constraints as barriers.
Seven Principles
The committee emphasized that trust is unnegotiable and should remain uncompromising. It highlighted people's first approach, adding that AI should strengthen human decision-making, but should postpone human judgment and citizen interest.
They called for the promotion of responsible innovation with purpose, equity and equity. The committee also emphasized that AI results should be fair and non-discriminatory. Moreover, it emphasized accountability, noting that responsibility lies on the entities deploying AI. AI said it should be understood by design to ensure the explanability of trust.
Finally, the committee emphasized safety, resilience and sustainability, saying that the AI system is safe, resilient and energy efficient.
Recommendations
The most important of the 26 recommendations in the panel's regulatory roadmap is the creation of a high-quality financial sector data infrastructure integrated with AI Kosh to democratize access to trustworthy AI models.
The committee also recommended that AI Innovation Sandboxes be installed to enable safe experimentation in the development of financial AI solutions. Share incentives, fundraising support, and shared “landing zones” to enable small, underserved financial institutions to adopt AI.
They called for the development of indigenous domain-specific AI models for the Indian financial sector, particularly those focusing on multilingual and inclusion.
In addition to embedding robust data lifecycle governance and AI model lifecycle monitoring, the committee also proposed adopting a step-by-step AI responsibility framework that provides oversight tolerance for first-time failures in safeguards, particularly for automated operational AI systems, but that keeps agencies responsible.
