The Reserve Bank of India (RBI) has cautioned banks and non-banking financial companies (NBFCs) against reliance on risk models based on pre-set algorithms emerging from collaborations with fintech companies, saying such models should be robust and their resilience tested periodically.
The central bank has advised these regulated entities to be careful of any undue risk building up in the system owing to information gaps which may lead to dilution of underwriting standards.
Click here to connect with us on WhatsApp
To mitigate biases in underwriting algorithms, the regulator emphasised the importance of ensuring that algorithms rely on accurate and diverse data sets. It highlighted the need to audit algorithms to identify minimum underwriting standards and potential discrimination factors.
“These models should be robust and their resilience should be tested periodically,” the RBI said in a report.
The regulator has highlighted that inherent biases and lack of transparency with artificial intelligence and machine learning (AI/ML) outcomes carry risks. It recommended lenders adopt ethical AI that can protect customer interest, promote transparency, security and privacy.
“It is, therefore, imperative to strike a balance between benefits and risks by strengthening the capacity of regulated entities (REs) and surveillance by oversight authorities, formulating/updating relevant legal and regulatory frameworks, proactively engaging stakeholders to identify possible risks, and expanding consumer education,” the regulator observed.
In another move to bring transparency within the fintech sector, the RBI earlier this month, announced its intention to establish a fintech repository by April next year. This repository aims to capture information about the activities, products, technology stack, and financial details of fintech companies.
More From This Section
It has undertaken the initiative to promote best practices within the fintech sector and enable regulators, among other stakeholders, to have relevant and timely information on fintech entities.
“Fintechs would be encouraged to provide relevant information voluntarily to the repository which will aid in designing appropriate policy approaches,” the report added.