The Black Box Banker: AI Credit Scoring & The Right to Know

In the traditional banking model, getting a loan involved sitting across a desk from a bank manager. If you were rejected, you could ask why. Perhaps your salary was too low, or you had a default five years ago. It was a human process.

In the FinTech era, the bank manager has been replaced by an algorithm. Hundreds of digital lending apps in India dispense instant credit by scraping your phone's data—your SMS logs, your location history, your contacts, and even your battery life. This "Alternative Data" is fed into a Black Box model which spits out a number: Yes or No. If the answer is No, there is no one to ask why. The algorithm has spoken.

Financial Inclusion vs. Algorithmic Exclusion

Proponents of AI credit scoring argue that it promotes "Financial Inclusion." By using alternative data, they can lend to the "credit invisible"—people with no formal credit history (CIBIL score) but who might be reliable borrowers (e.g., a street vendor who pays suppliers digitially).

However, the flip side is "Algorithmic Exclusion." What if the AI decides that people who charge their phone at 2 PM are irresponsible borrowers? What if it redlines entire neighborhoods because GPS data defines them as "high risk"? These correlations might be mathematically statistically significant but legally discriminatory.

The Legal Framework: RBI's Digital Lending Guidelines

The Reserve Bank of India (RBI) has been proactive. Its Digital Lending Guidelines (2022) explicitly prohibit apps from accessing mobile phone resources like file systems, media, and contact lists. This was a direct response to predatory lending apps that used contact lists to harass borrowers.

But regarding the AI models themselves, the regulation operates on a "Light Touch" basis. There is no requirement for "Explainability of Credit Decisions." This is a significant gap. Under the Fair Credit Reporting Act (FCRA) in the US, denied borrowers have a right to "Adverse Action Notices"—specific reasons for rejection. Indian borrowers currently lack this statutory right against FinTechs.

The Problem of Proxy Discrimination

The Constitution of India prohibits discrimination by the State. Private banks are not "State," but they serve a public function. If a FinTech's algorithm charges higher interest rates to women or minorities, is it illegal?

Currently, private sector discrimination is largely unregulated in India outside of specific employment contexts. AI allows for "Proxy Discrimination." An algorithm might not use "Religion" as a variable, but if it uses "Meat Consumption" (derived from grocery delivery data) as a proxy for risk, it effectively penalizes certain communities. This technological loophole allows for automated prejudice that is invisible to current regulators.

The Data Privacy Angle

The DPDP Act, 2023 offers a ray of hope. It mandates that data collection must be necessary for the purpose. Is access to my Facebook friends list necessary to determine my ability to repay a ₹50,000 loan? Arguably not.

We are advising FinTech clients to embrace "Data Minimization." The era of "grab all data now, find a use later" is over. Every data point fed into a credit model must have a justifiable link to creditworthiness.

The Future: 'Fair Lending' Audits

To build trust, the Indian FinTech sector must move towards voluntary "Fair Lending" audits. Third-party auditors should test models for disparate impact.

Imagine a future where your digital loan app comes with a "Fairness Certified" badge. This would signal to consumers that while the decision is automated, it is not arbitrary.

Conclusion

Credit is the lifeblood of economic mobility. Determine who gets credit, and you determine who gets to buy a house, start a business, or send their child to college. We cannot cede this immense power to un-auditable black boxes. The Right to Credit is not a fundamental right, but the Right to Fair Treatment in the financial marketplace must be defended.