Whitepaper: AI Governance for Lending in India: Components & Challenges

As and automation technology mature, the need for inherently interpretable, explainable and responsible models has become the critical focus. While this development is being encouraged, there has been an increased emphasis on managing associated risks with these technologies. The AI/ ML regulatory landscape in India is changing rapidly; it has become imperative for organizations to make requisite tweaks in their business processes and explain to regulators how their system works to demonstrate compliance with applicable regulations.

The Reserve Bank of India recently firmed up a regulatory framework to support the orderly growth of credit delivery through digital lending methods while mitigating regulatory concerns. One of the many recommendations accepted for immediate implementation include:

i) REs (Regulated Entities) to ensure that the algorithm used for underwriting is based on extensive, accurate and diverse data to rule out any prejudices. Further, the algorithm should be auditable to point out minimum underwriting standards and potential discrimination factors used in determining credit availability and pricing.

ii) Digital lenders should adopt ethical AI, which focuses on protecting customer interest, promotes transparency, inclusion, impartiality, responsibility, reliability, security and privacy.

_{*As mentioned in RBI press release on 'Recommendations of the Working group on Digital Lending - Implementation' (link)}

Organizations need to be aware of the regulatory and ethical implications of the guidelines on business.

This whitepaper, based on the proceedings of our workshop 'Responsible AI/ ML in Lending'. , will guide regulators and all stakeholders depending on AI decisions about the impact of RBI guidelines, components required for AI Governance, and the challenges, threats and opportunities while building such AI Governance.

‍

Contents:

• Introduction and components required for ‘AI Governance’
• Designing an AI Governance framework:

    - Reliability

    -  Fairness and Bias

    - ML Explainability

    - Algorithmic Auditing

    - Data Privacy

     - Responsible AI

     - AI usage risk