HomeNewsOpinionWhat makes the RBI concerned about algo model-based lending

What makes the RBI concerned about algo model-based lending

The apprehensions surrounding algorithmic lending must be contextualised within the broader landscape of financial innovation. The increasing reliance on newer datasets including unstructured data, and digital technologies is an inexorable shift, driven by the need for efficiency, speed and adaptability in financial services

December 05, 2023 / 16:44 IST
Story continues below Advertisement
Lending,
A lack of transparency raises valid concerns about accountability and fairness in lending practices.

Resonating with a cautionary tone, RBI governor Shaktikanta Das recently articulated concerns over algorithmic model-based lending. This reiterates the regulatory commitment to maintaining the stability and resilience of the financial sector in the face of evolving fintech dynamics. Algorithmic models, powered by artificial intelligence and machine learning, often operate as ‘black boxes’, making it challenging to decipher the rationale behind their decisions. This lack of transparency raises valid concerns about accountability and fairness in lending practices. Many lending models are built using algorithms. The fear is that these algorithms, if not rigorously monitored, and if not traceable to root cause, could inadvertently perpetuate biases or discriminate against certain variables and also lead to bad credit decisions.

In response to potential risks, the banking regulator recently raised the risk weights for consumer credit exposure by 25 percentage points to 125 percent. This reflects concerns about the possibility of undue risk accumulation in the system arising from information gaps in these models, potentially resulting in a dilution of underwriting standards. The RBI has been kind and patient in asking the boards of its regulated entities to be vigilant, and to take accountability about lending models and lending exposure across segments. To be fair to the RBI, these are essential roles of the individual boards to ensure prudential and risk-managed behaviour in their enterprises.

Story continues below Advertisement

Updating Algorithms Dynamically

To be fair to those financial institutions that are risk-prudent, they revamp their credit models every 12-18 months and update their algorithms dynamically. Regulators should make this a rule, instead of worrying about the use of algorithmic models. In reality, algorithm-based lending has not only reduced operational costs but has also expanded the outreach of financial services providers. The collaboration between banks, NBFCs and fintechs has increased, leading to the introduction of innovative products, services and business models.