AI Lending Expands Credit Access, but Oversight Matters

Many financial services are, at their core, information businesses. For example, an insurance company needs to have correct information about the future rate of accidents and their cost to properly identify the price at which an insurance policy will still be profitable.

Similarly, lending requires properly assessing the risk of default from a given customer, and setting the interest charged at the appropriate level.

Historically, this was first done with sole human judgement and primitive mathematical formulas, with social class, personal connections, and personal behaviour impacting an individual’s access to financial services.

More professional and objective methods progressively emerged to properly assess risk with less bias. For example, credit scores can become, in America, the dominant form of evaluation for assessing an applicant’s borrowing capacity.

But even more advanced methods are now becoming increasingly prevalent, using AI to assess lending risks with even more precision, taking advantage of the treasure trove of useful data not used by traditional credit scores.

So AI lending can reach borrowers normally neglected by conventional credit models, and overall improve risk assessment. However, this broadening of access to credit can also carry risks if done improperly, from mistakes in actual underwriting quality, increasing discrimination, etc.

A new study by researchers at Kunming University (China),  Shanghai University of Finance and Economics (China), and University of Hamburg (Germany) investigates the effects of artificial intelligence in credit markets.

It found AI lending raises financial inclusion by 8.4–11.2 percentage points on average, mostly for FinTech lenders. But it also found that AI-driven credit expansion amplifies tail-risk volatility.

They published their findings in International Review of Economics & Finance¹, under the title “Artificial intelligence in credit markets: Algorithmic lending, financial inclusion, and regulatory arbitrage”.

Analyzing AI Lending

AI Lending Promises And Risks

From a novelty adopted by newcomers like Upstart (see below for more on this company), AI-related technologies are now present in most lending processes in some capacity.

“Machine learning algorithms now underpin loan origination systems at major retail banks, digital lending platforms, and a growing constellation of FinTech entrants.”

These systems can process and calculate thousands of variables spanning transactional histories, psychometric assessments, mobile behavioral patterns, and real-time social signals. But this is not without causing potential new problems:

• They may perpetuate or amplify historical biases embedded in training data
• They operate with a degree of opacity that frustrates supervisory oversight;
• Their deployment across jurisdictional boundaries invites a form of regulatory arbitrage that challenges the territorial logic of prudential regulation.

Building The Right Data

To understand not just the effects of AI lending at the local level, but also cross-jurisdictionally, the researchers analyzed 42 economies drawn from Asia-Pacific (14 countries), Europe (10 countries), Latin America (10 countries), and Sub-Saharan Africa (8 countries), over the 2012–2022 period.

They extracted data for these countries on financial inclusion, supervisory architecture quality, and macro-prudential framework comprehensiveness.

The most important metric, financial inclusion, was measured by taking into account:

• Formal credit account ownership rate.
• MSME credit penetration.
• Borrower-to-adult population ratio.
• Number of commercial bank branches per 100,000 adults.
• Mobile money account ownership rate.

They also created a new novel indicator of AI lending adoption intensity (ALII) to assess how much AI lending is prevalent in a given country.

“The ALII functions as a structural metric of macro-level technology adoption intensity and sector-wide capacity distribution. Rather than auditing individual model accuracy or algorithmic alignment at the micro-level, the index quantifies the structural density and operational penetration of automated decision-making pipelines within the broader financial ecosystem.”

Study Results

Great Disparities

Unsurprisingly, the study found that financial inclusion varies greatly from one region to the other, in a range from 0.042 in low-income Sub-Saharan African economies in early sample years to 0.941 in high-income Asia-Pacific economies in later years.

The AI Lending Intensity Index displays even greater dispersion: the mean value of 0.317 masks a coefficient of variation of 0.89, with low-income economies averaging 0.089 and high-income economies averaging 0.524.

Another finding is that there was a notable acceleration of ALII after 2016, which coincides with the commercial deployment of deep learning applications in credit scoring. Meanwhile, the highest values for Credit Cycle Volatility (CCV) were observed in emerging market economies in the post-2015 period, coinciding with rapid FinTech credit expansion documented in the ALII data.

AI Lending Drives Inclusion

The study found a direct link between AI lending and financial inclusion.

“A one standard deviation increase in the AI Lending Intensity Index is associated with an increase in the Financial Inclusion Index of 0.084 standard deviations.”

It also appeared that the positive effect of AI lending on financial inclusion is stronger when supervisory capacity is weaker. This could be due to two possible and non-mutually exclusive reasons:

• Weaker supervisory environments contain larger pools of underserved borrowers whom AI lenders can profitably serve using alternative data.
• AI lenders in low-supervision environments are more willing to extend credit to marginal borrowers because the regulatory costs of potential default attribution are lower.

This effect also seems to be driven principally by FinTech companies, as Tech AI lenders generate substantially larger financial inclusion dividends (3x) than incumbent banks deploying AI tools.

“One standard deviation increase in FinTech AI lending intensity raises FII by approximately 13.8 percentage points, nearly three times the effect attributable to equivalent AI adoption by incumbent banks.”

So it seems that increasing financial inclusion is not linked to AI reducing processing costs for existing bank customers, but to FinTech using AI to reach borrowers excluded from the conventional banking system.

This fits a general view that established banks deploy AI defensively (protecting existing customer relationships) rather than expansively (targeting new segments).

Greater Risks Depending On Countries?

The study found that increased financial inclusion through AI lending is not significantly tied to higher systemic risks in low-supervision environments. So it is likely that this is the setup where AI lending will produce the best results, with greater increases in financial inclusion and lower risks.

In contrast, in more developed economies and more tightly regulated financial environments, AI lending generates more moderate inclusion benefits. In that case, it is also causing a small, but non-negligible increase in risks. This is likely because only the most marginal borrowers are truly excluded from any type of lending in these economies.

“Financial inclusion is already near-universal among the adult population, leaving little scope for AI lending to expand the margin of included borrowers.”

Like with all lending, the greater risks are not during normal market conditions, but during periods of high stress in the financial system.  So while automated lending seems stable during quiet times, it triggers rapid, uniform capital cuts during crises, making solid macro-prudential buffers vital for stopping systemic crashes and market panic.

Regulatory Arbitrage

A slightly more concerning discovery of the study is the tendency of FinTech companies using AI for lending to try to escape tighter regulations, including through the use of shell companies.

“In the Asia-Pacific region, as local supervisory intensity climbed from ＋0.02 to ＋0.22, firms did not slow down. Instead, they moved their operations off-grid, jumping from 2 to 11 off-perimeter shell subsidiaries. At the same time, they more than tripled their cross-border asset routing from 14.5% to 52.8%, prompting a spike in legal disputes.”

It indicates that local regulations are struggling to fully regulate this type of lender. Instead, it encourages agile digital lenders to unbundle their businesses and push capital into unmonitored territories.

Investors Takeaway On AI Lending

This study confirms previous ones that showed that AI lending improves financial inclusion. For example, it was already known that algorithmic mortgage underwriting in the United States increases approval rates for minority borrowers without commensurate increases in default, a result consistent with discrimination-based exclusion rather than risk-adjusted exclusion.

It now demonstrates that this effect is consistent in virtually every region and economy type, as well as supporting a direct causal link between increased AI lending and better and fairer access to capital. This effect is especially strong in regions with poor financial inclusion and/or weaker supervisory environments.

At the same time, the study also found that higher risks and fewer benefits are common for developed economies.

So it could mean that while initially AI lending increased access to capital, the greatest benefit and lowest risks could be found in frontier markets. With, of course, the caveat that such markets are generally inherently riskier than developed economies, independently of the quality of an individual AI lending company.

Lastly, it seems that FinTech companies are quick to exploit regional arbitrage and try to avoid tighter regulations. This can be a long-term risk, as improved regional coordination or a harder response by regulators could impact margins and even lead to legal cases for these companies, a risk to consider for their shareholders.

So overall, investors in Fintech companies engaged in AI lending should assess whether these platforms can expand access sustainably, alongside their loan growth.

If there is a large enough pool of people poorly served by the existing financial industry, AI lending is likely to be of great benefit with a large addressable market.

If the pool is too small, increased risks of default during market stress or equally risky cross-jurisdiction methods to avoid regulations can carry unseen dangers for the company’s shareholders in the long term.

Investing In AI-Driven Lending

Upstart Holdings

As lending and credit scores evolve, new companies are emerging to provide better alternatives to traditional credit scores and lending practices. One of them is Upstart, a lending marketplace powered by AI, launched in 2012, much before AI became a center of tech conversation.

The idea behind Upstart’s creation was that the existing credit score system is inefficient and outdated. With a lot more data available, it had to be possible to identify loan risks better and, as a result, provide cheaper loans to a large portion of the population.

“Our personal loan model is meaningfully more accurate than a traditional credit model at evaluating borrower risk. We believe years of modeling improvements lie ahead.”

This early start gave the company an abundant source of data on which to train its own AI and leverage the recent progress made in AI training methods.

Source: Upstart

This means Upstart’s method can identify people with high FICO scores but, in practice, have a high risk of defaulting on their loans. And conversely, people with low FICO scores are not that likely to default.

So with the use of AI and data, Upstart can achieve outstanding results for a given cohort. For example, it can manage to offer more loans for the same level of risk, rising from a 27% approval rate to 67%. Or it can reduce the losses significantly, from a 9% to 5% default rate, or almost half as much.

Source: Upstart

In addition, using an algorithm instead of a loan officer means the process is automated. This is both less expensive to do and much quicker and more pleasant to apply for a loan.

Upstart’s marketplace connects borrowers with banks, credit unions, and institutional capital. It reported that institutional investors purchased 64% of the principal originated through its marketplace in 2025.

Upstart serves an absolutely massive market, with $25T in credit originated each year globally. Of this sum, up to $1T in revenue is available to the platforms that originate and service that credit.

In 2025, the company originated 1.5 million loans (up 115% year-to-year), for a total of $11B (up 86% year-to-year), and total revenues for the company of $1B (up 64% year-to-year). The company is also growing quickly in new segments, with, for example, auto loan origination growing 5x (from $45M in Q4 2024 to $263M in Q1 2026), and home loan origination growing 6x (from $27M in Q4 2024 to $143M in Q1 2026).

So even if already large, the current number of loans pales in comparison to the full potential of advanced credit score methods and AI-driven loan assessment.

Source: Upstart

The recent growth numbers and having been built as an AI-first company are strong positives for the future of Upstart, even if the stock price has not really rewarded this success, with prices at the same level as in 2022, and much below its 2021 peak during the pandemic.

In part, this is because there are also a few risks to Upstart’s business model.

The first is that its model can deteriorate, as economic conditions move beyond training data and its AI could miss new trends.

Another issue is increasing regulatory pressure for explainability, fairness, and demographic performance reporting, as even if inadvertently, AI could bring back discriminatory practices previously banned from the financial industry.

Lastly, the company still remains dependent on institutional funding remaining available across credit cycles, with signs of stress in the financial system building in the context of high debt levels and a tense geopolitical situation.

Still, Upstart is ideally placed to become a key AI lending marketplace and underwriting infrastructure provider, a necessary evolution of the initial FinTech and P2P networks’ push for new lending practices.

And investors might want to get exposure to a company that has been radically changing how loan underwriting is performed early on.

Latest Upstart Holdings  (UPST) Stock News and Developments

Study Referenced

^{1. Podong Song, Shutong Wu, Janson Partab. Artificial intelligence in credit markets: Algorithmic lending, financial inclusion, and regulatory arbitrage. International Review of Economics & Finance. December 2026. Article: 105937. Volume 112. 10.1016/j.iref.2026.105937}