How Algorithmic Lending Is Reshaping Credit Access for India's Underbanked

The same basic technology — an algorithm reading your phone's data to decide whether you're creditworthy — has produced two almost opposite stories in India over the past few years. In one version, it's quietly solving a problem India has struggled with for decades: how do you extend credit to someone with no formal income proof, no collateral, and no credit history? In the other version, it's harvesting a borrower's entire contact list to blackmail them into repaying a seven-day loan at an effective annual rate north of 200%, sometimes with tragic results. Both versions are real, running in parallel, built on the same underlying capability. Understanding algorithmic lending in India means understanding why the technology split into these two paths, and what's actually being done about the second one.

The genuine inclusion story

India's traditional credit system was built around documentation most of the country's population never had: formal payslips, income tax returns, a multi-year credit bureau history. That structurally excluded informal workers, gig economy earners, first-time borrowers, and small businesses running mostly in cash — which is to say, a very large share of the country. Alternative credit scoring changes the inputs entirely. Instead of a CIBIL score built on formal loan history, lenders now assess UPI transaction patterns, mobile recharge regularity, utility bill payments, GST filings for small businesses, and e-commerce behavior — signals that exist for people who've never held a formal loan but do use a smartphone and make digital payments regularly.

The performance gains are measurable, not just directional. Academic modeling on Indian lending data found AI-powered alternative scoring models achieving AUC-ROC prediction scores above 0.85, compared to roughly 0.72 for traditional CIBIL-based approaches — with alternative data reducing default-prediction error by an estimated 18-22%. On the small business side, NBFCs and fintechs now routinely combine transaction histories, GST filings, and utility payments to assess MSME creditworthiness in ways conventional, documentation-heavy underwriting never could, and government-backed platforms like PSB Loans in 59 Minutes demonstrate what near-instant, AI-assisted approval looks like when it's built by regulated entities under proper oversight. This is a real, well-evidenced improvement over the alternative — which, for most of the underbanked population, wasn't a worse loan product, it was no loan product at all, or a moneylender charging considerably worse terms with none of the disclosure this technology at least makes possible.

The same mechanism, weaponized

Here's the uncomfortable part: the exact same technical capability — extracting rich behavioral signal from a smartphone to make a fast lending decision — is also what made India's predatory loan app crisis possible at the scale it reached. Illegal lending apps, many later traced to entities with links outside India, used the permission-request stage of app installation not to build a better credit model, but to harvest a borrower's entire contact list and photo gallery. When repayment on a short-term loan — often structured with an unrealistic seven-day term specifically engineered to trigger default — came due, the harassment machinery activated: threatening calls to the borrower's family and colleagues, morphed and doctored photos sent to their entire contact list, and psychological pressure campaigns designed to shame borrowers into paying far more than they legally owed. India's Home Ministry, in a directive to state governments, explicitly linked this pattern to "multiple suicides by citizens owing to harassment, blackmail, and harsh recovery methods" — language serious enough that it prompted a sustained, multi-year government crackdown rather than routine regulatory tightening.

The scale of the enforcement response gives some sense of how large this problem became: Google removed more than 2,000 apps from the Play Store, the Enforcement Directorate seized over ₹1,000 crore in app-based loan fraud assets between 2023 and 2025, and government crackdowns targeting apps with alleged foreign financial links have continued into 2026, with hundreds more apps blocked and new lists of banned platforms published regularly as bad actors relaunch under new names to evade detection.

What changed in the RBI's 2026 rulebook

The regulatory response has been substantial, and it's worth being specific about what actually changed, because it reframes what "algorithmic lending" is now legally allowed to look like in India. Under the updated 2026 Master Direction on Digital Lending, only Reserve Bank-regulated entities — scheduled banks, small finance banks, or NBFCs — or their properly authorized Lending Service Provider agents can legally operate a digital lending app at all, closing the loophole that let unregistered shell operations pose as fintechs. Contact and call-log access is now explicitly prohibited — any app requesting "Read Contacts" permission is, by definition, non-compliant and reportable. Contact scraping for harassment purposes is now a criminal offense under the DPDP Act, not merely a regulatory violation. Lenders must issue a legally binding Key Fact Statement disclosing the true annual percentage rate before any money changes hands, replacing the old practice of advertising a low headline rate while burying the real cost in fees and penalties. A mandatory three-day cooling-off period gives borrowers a chance to walk away, and First Loss Default Guarantee arrangements — where a tech platform effectively fronted risk for a bank without real skin in the game — are now capped at 5%, forcing regulated lenders to retain genuine underwriting responsibility rather than outsourcing risk entirely to an intermediary with no license to lose.

The market-level effect, per industry analysis, has been a genuine consolidation: unregulated, unlicensed apps are exiting, while compliant platforms — the market now serves over 40 million borrowers across a sector exceeding ₹5.2 lakh crore — are gaining borrower trust as the bad actors get filtered out. This is a case where tighter regulation, rather than simply raising compliance costs the way it might in a mature market, appears to be actively rebuilding the credibility a technology needs to actually deliver on its stated inclusion goal. A market where borrowers reasonably fear that any given loan app might photoshop their face onto something obscene and mail it to their relatives is not a market that expands financial inclusion, no matter how sophisticated its underwriting algorithm is — trust is a precondition for adoption, not an afterthought.

The lesson this holds for algorithmic systems generally

This connects to a pattern this series has traced in other contexts — the "algorithmic manager" governing gig workers, or the opacity concerns around AI-driven welfare authentication. The technology itself doesn't determine the outcome; the business model and oversight regime wrapped around it does. An algorithm reading UPI transaction history to extend fair credit to a previously unbanked worker, and an algorithm reading a phone's entire contact list to build a blackmail target list, are using structurally similar data-extraction techniques toward completely opposite ends. The difference between India's genuine fintech-driven financial inclusion story and its predatory lending crisis wasn't the sophistication of the AI — it was whether a regulator existed with the will and the tools to distinguish a legitimate credit decision from a data-harvesting operation dressed up as one, and to make the second one illegal rather than merely distasteful.

India's 2026 framework suggests the country has, belatedly, built that distinction into law. Whether it holds — as banned apps keep relaunching under new names and new Play Store listings — will determine whether algorithmic lending's actual legacy in India ends up being the inclusion story or the harassment story it's currently telling simultaneously.

This piece touches on documented cases involving suicide linked to loan app harassment; if you or someone you know is struggling, please reach out to a mental health professional or a crisis helpline in your area.