1. Introduction

Alternative lending increasingly relies on artificial intelligence (AI), machine learning, automated underwriting, alternative-data scoring, and algorithmic pricing. Fintech lenders may use transaction histories, mobile-payment data, bank-account information, employment data, behavioural indicators, platform activity, device information and other non-traditional variables to determine:

• whether a borrower receives credit;
• the amount of credit;
• interest rates and fees;
• credit limits;
• repayment periods;
• risk classifications;
• collection strategies; and
• dynamic loan pricing.

These technologies can improve credit assessment and reduce underwriting costs. However, competition concerns arise when competing lenders use common AI systems, shared alternative data, common vendors, coordinated pricing parameters, or algorithms capable of learning from competitors' behaviour.

The central competition-law problem is:

When does independent algorithmic decision-making become coordination between competing lenders?

The distinction is important. An AI system independently learning from publicly available market conditions is not automatically a cartel. The legal risk becomes substantially greater where competing lenders exchange competitively sensitive information, jointly design pricing parameters, use a common intermediary to coordinate their decisions, or knowingly delegate important competitive decisions to an algorithm designed to align them.

2. Meaning of Alternative Lending AI Coordination

Alternative lending refers broadly to credit supplied outside conventional bank lending structures, including:

• peer-to-peer lending;
• marketplace lending;
• buy-now-pay-later platforms;
• embedded finance;
• digital consumer loans;
• fintech business lending;
• invoice-financing platforms;
• app-based microcredit;
• platform-based SME lending; and
• AI-driven private credit.

AI coordination occurs where algorithms used by competing lenders have the effect—or are allegedly designed—to reduce independent competitive decision-making.

For example:

Lender A + Lender B + Lender C → common AI vendor → common data pool → pricing recommendation → similar interest rates

This structure can create a hub-and-spoke coordination risk.

The fact that the coordination occurs through software rather than a telephone call does not necessarily remove it from competition law.

3. Principal Forms of AI Coordination Risk

A. Common AI Pricing Algorithms

Several competing lenders may subscribe to the same AI pricing system.

The system may recommend:

• interest rates;
• risk premiums;
• origination fees;
• credit limits;
• late-payment charges; or
• borrower-specific pricing.

If the algorithm uses confidential information supplied by competing lenders, the system may reduce the lenders' ability to make genuinely independent pricing decisions.

The concern is particularly serious where the system incorporates:

• competitors' current rates;
• loan volumes;
• default rates;
• borrower conversion rates;
• planned pricing changes;
• risk models; or
• future lending strategies.

4. Competitively Sensitive Alternative Data

Alternative lending creates an additional information-exchange problem.

Suppose competing fintech lenders provide a common AI provider with:

• borrower default probabilities;
• application acceptance rates;
• loan pricing;
• customer acquisition costs;
• credit limits;
• borrower churn;
• delinquency data; and
• future pricing strategies.

The AI provider could theoretically aggregate these datasets and generate recommendations for competing lenders.

The competition concern is not simply the existence of data. It is what information is shared, how current it is, whether it is identifiable, and how it influences competitive decisions.

The CJEU has treated exchanges of future credit-pricing intentions and risk variables as particularly sensitive. In a recent banking-information case, the Court held that confidential information concerning future credit spreads and risk variables could constitute coordination restricting competition by object.

5. AI-Based Interest-Rate Coordination

Interest rates are one of the most competitively sensitive variables in lending.

An AI system could continuously monitor:

• competitors' loan rates;
• central-bank rates;
• borrower demand;
• approval rates;
• competitor market shares;
• loan maturities; and
• competitors' responses to pricing changes.

The algorithm could then recommend a rate that avoids aggressive price competition.

The problem becomes especially acute in concentrated lending markets.

Example

Suppose five alternative lenders compete for SME borrowers.

If one lender reduces its interest rate from 14% to 12%, the AI systems of the other lenders immediately detect the change and automatically respond.

If all five systems have been programmed to maintain a particular pricing corridor, competition may become significantly weaker.

6. Algorithmic Parallelism Versus Illegal Coordination

A crucial legal distinction must be maintained.

Mere parallel behaviour

Five lenders independently use AI systems and independently arrive at similar prices.

This does not automatically establish a cartel.

Coordination

Five lenders:

1. provide confidential pricing data to the same AI intermediary;
2. know that competitors are participating;
3. agree to rely on the resulting recommendations; and
4. use those recommendations to maintain aligned pricing.

This creates a considerably stronger competition-law concern.

Therefore:

Algorithmic price similarity is evidence that may require investigation; it is not by itself proof of unlawful agreement.

7. Hub-and-Spoke AI Lending

A particularly important model is:

Lender A
↓
AI Platform
↑
Lender B
↑
Lender C

The AI platform becomes the hub, while the competing lenders are the spokes.

Potentially problematic conduct includes:

• exchanging non-public lending rates;
• sharing future pricing strategies;
• sharing individual competitor loan data;
• agreeing to use the same pricing recommendations;
• communicating acceptance of competitor participation;
• incorporating competitors' confidential data into AI models.

The U.S. RealPage litigation illustrates how antitrust authorities can treat common algorithmic intermediaries and competitor data-sharing as potential coordination. Although it concerned rental pricing rather than lending, the structural reasoning is relevant to AI-based alternative lending. The DOJ alleged that competing businesses supplied non-public information to a common algorithmic pricing system and received recommendations based upon that collective information.

8. Six Important Case Laws

Because there are still relatively few reported judicial decisions specifically concerning AI coordination between alternative lenders, established algorithmic-collusion, information-exchange, hub-and-spoke and credit-market cases provide the principal legal analogies.

1. United States v. Topkins

United States v. Topkins is one of the clearest algorithmic-pricing precedents.

The defendant and other online sellers discussed prices and agreed to use pricing algorithms to implement the arrangement.

The important principle is:

Using an algorithm to implement an existing agreement does not make the underlying price-fixing lawful.

For alternative lenders, the analogy is direct.

If competing lenders agree on lending prices or pricing parameters and then instruct an AI system to implement the agreement, the AI is merely the technological mechanism through which the cartel operates.

The algorithm therefore does not provide a defence to traditional cartel liability.

Relevance to lending

Potential examples include agreements concerning:

• minimum interest rates;
• maximum discounts;
• common risk premiums;
• minimum origination fees; or
• common BNPL charges.

2. Eturas v. Lietuvos Respublikos konkurencijos taryba

Eturas is a leading European algorithmic-coordination case.

The case involved an electronic platform through which travel agencies operated. A common electronic system imposed a restriction affecting discounts.

The CJEU considered when awareness of a platform-wide restriction, together with subsequent market behaviour, could establish participation in a concerted practice.

The case is important because competition law can apply to coordination conducted through electronic systems rather than traditional face-to-face communications.

For alternative lending, the analogy is a common lending platform or AI intermediary communicating a pricing restriction or other competitively sensitive parameter to competing lenders.

However, mere receipt of an automated message is not necessarily sufficient by itself; the circumstances surrounding knowledge, participation and subsequent conduct matter.

3. United States v. Container Corporation of America

United States v. Container Corp. of America, 393 U.S. 333 (1969) is a foundational information-exchange case.

Competitors exchanged current pricing information with one another. The Supreme Court found sufficient concerted action and concluded that the information exchange had an anticompetitive effect in the circumstances of that market.

Relevance to AI lending

Imagine competing digital lenders providing a common AI system with:

• current interest rates;
• borrower-specific prices;
• loan terms;
• credit-risk premiums; and
• pricing responses.

The technological form of the information exchange would not necessarily eliminate the underlying competition concern.

Container Corporation therefore establishes an important principle:

Information concerning competitors' current pricing can itself become competitively significant when exchanged through a coordinated arrangement.

4. T-Mobile Netherlands BV v. Raad van bestuur van de Nederlandse Mededingingsautoriteit

The T-Mobile Netherlands case concerned exchange of strategically important information between competitors.

The CJEU emphasized that information exchange can constitute a restriction of competition where the information reduces uncertainty concerning competitors' future market conduct.

The case is particularly relevant to AI because algorithms function most effectively when supplied with accurate information concerning competitors' likely future decisions.

Lending application

Suppose lenders exchange information concerning:

• planned interest-rate changes;
• future credit standards;
• expected risk premiums;
• intended loan volumes; or
• planned tightening of underwriting criteria.

Feeding such information into a common AI system could reduce strategic uncertainty between competing lenders.

The legal issue therefore is not merely "Was data exchanged?", but:

Did the exchange remove commercially important uncertainty between competitors?

5. RealPage Antitrust Litigation

United States and State Plaintiffs v. RealPage, Inc. is one of the most important contemporary algorithmic-coordination matters.

The DOJ alleged that competing landlords supplied non-public, competitively sensitive information to RealPage's algorithmic pricing system and that the system generated pricing recommendations using that collective information.

Although RealPage concerns rental housing rather than lending, the underlying structure is highly relevant:

Competitors → common software provider → confidential competitor data → algorithmic recommendation → competitor pricing

The DOJ subsequently obtained a proposed settlement requiring RealPage to stop using competitors' non-public competitively sensitive information for runtime rental-price recommendations and imposing restrictions concerning model training.

Lending analogy

Replace:

landlords → lenders

rents → interest rates

occupancy data → borrower/loan data

rent recommendations → loan-price recommendations

The resulting competition-law problem is conceptually similar.

6. In re RealPage, Inc. Rental Software Antitrust Litigation

The private RealPage litigation provides another important judicial development.

Courts have considered whether allegations involving:

• common algorithmic software;
• competitors' confidential information;
• parallel pricing;
• adoption of algorithmic recommendations; and
• plus factors

are sufficient to plead an unlawful conspiracy.

The litigation is particularly relevant because algorithmic coordination creates a difficult evidentiary problem: parallel outcomes may arise either from independent optimisation or from coordinated conduct.

In the multifamily case, the court allowed significant allegations concerning alleged algorithmic coordination to proceed, while other allegations were treated differently. The litigation therefore demonstrates the importance of factual evidence surrounding the algorithm, data inputs and participants' conduct rather than simply observing similar prices.

9. Additional Relevant Case: Apple Inc. v. Pepper

Apple Inc. v. Pepper is not an algorithmic-collusion case, but it is useful when analysing digital-platform intermediation.

It illustrates how platform structures can create competition-law questions concerning the relationship between:

• platform operators;
• suppliers;
• customers;
• commissions; and
• market access.

For alternative lending platforms, a similar issue arises where an AI-powered marketplace controls access between borrowers and multiple lenders.

The competition analysis can therefore extend beyond price coordination to platform power and intermediary control.

10. AI Coordination Through Common Data Pools

Alternative lending platforms frequently rely on large data pools.

A common AI provider may collect:

The more current, granular, individualised and forward-looking the information, the greater the potential coordination concern.

11. Tacit AI Coordination

A more difficult problem arises when there is no express agreement.

Consider:

• Lender A's algorithm observes Lender B.
• Lender B's algorithm observes Lender C.
• Lender C's algorithm observes Lender A.
• Each system learns that aggressive price reductions trigger retaliation.
• The algorithms gradually settle on stable prices.

No human executive expressly agrees to fix prices.

This produces a distinction between:

A. Agreement-based coordination

There is evidence of:

• communication;
• common instructions;
• data exchange;
• agreement to use the same system; or
• acceptance of a common pricing rule.

B. Autonomous algorithmic parallelism

Each lender independently deploys AI and the algorithms independently converge on similar pricing.

The second situation creates a difficult competition-law problem because many antitrust regimes traditionally require an agreement, concerted practice or other legally cognisable coordination.

Academic and legal analysis continues to distinguish these situations rather than treating every algorithmically aligned outcome as cartel conduct.

12. Machine Learning Creates a Special Problem

Traditional software generally follows predetermined instructions.

Machine-learning systems may instead:

1. receive historical data;
2. identify correlations;
3. update predictions;
4. optimise objectives;
5. respond to market conditions; and
6. change recommendations over time.

Consequently, an AI lending system may produce coordination without an explicit human instruction to coordinate.

For competition law, this raises several questions:

• Who designed the objective function?
• What data was used for training?
• Was competitor data included?
• Who selected the variables?
• Were competitors informed about the system?
• Did lenders know that competitors were using the same system?
• Could lenders reject the AI recommendation?
• Did the system penalise aggressive price competition?
• Did the provider communicate competitor information?
• Were lenders contractually required to use particular recommendations?

13. The "Black Box" Problem

AI coordination investigations can be complicated because the algorithm may be difficult to explain.

Competition authorities may need to examine:

• source code;
• training datasets;
• model architecture;
• optimisation functions;
• feature-selection processes;
• logs;
• API calls;
• pricing outputs;
• model updates;
• communications between customers and the AI provider; and
• evidence of human intervention.

A lender cannot necessarily avoid competition scrutiny simply by saying:

"The AI made the decision."

The relevant legal question may instead be:

What human and organisational decisions caused the AI to behave in this manner?

14. Common-Vendor Risk

A particularly important risk exists where most alternative lenders purchase AI services from the same provider.

Potentially legitimate structure

Each lender provides only its own historical data, and the provider supplies an independently developed model.

Higher-risk structure

Each lender provides:

• current rates;
• current borrower acceptance rates;
• future pricing;
• credit limits;
• competitor strategy;

and the provider combines the data to generate recommendations for all participating lenders.

The second model creates much greater information-exchange and coordination risk.

15. Algorithmic Credit-Risk Coordination

Coordination need not concern interest rates alone.

Competing lenders could potentially align:

• credit-score thresholds;
• loan-to-income limits;
• debt-service ratios;
• minimum income requirements;
• acceptable default probabilities;
• collateral requirements;
• loan duration;
• rejection thresholds.

For example:

Lender A: reject borrowers above 8% predicted default probability
Lender B: reject borrowers above 8%
Lender C: reject borrowers above 8%

If these thresholds independently result from similar risk models, that is not necessarily unlawful.

But if competing lenders exchange confidential underwriting parameters and agree to adopt a common AI-generated threshold, the competition-law concern is materially different.

16. Coordinated Market Allocation

AI can also facilitate allocation of borrowers.

For example, a lending platform might use an algorithm that determines:

• which lender receives which borrower;
• which lender receives high-risk customers;
• geographic territories;
• SME sectors;
• borrower categories;
• loan sizes.

If competing lenders agree through the platform not to compete for particular categories of borrowers, this could raise market-allocation concerns.

17. Bid-Rigging in Digital Lending

Alternative lending platforms may also facilitate financing for:

• infrastructure;
• real-estate projects;
• corporate acquisitions;
• renewable-energy projects;
• SME financing.

Where several lenders compete to finance the same transaction, AI systems could theoretically coordinate:

• financing rates;
• fees;
• loan amounts;
• security requirements;
• maturity;
• underwriting conditions.

An agreement to rotate or allocate financing opportunities could create a cartel concern independent of whether AI was used.

18. Predatory or Exclusionary AI Coordination

Competition concerns can also arise under abuse-of-dominance principles.

A dominant AI lending platform might:

• deny rival lenders access to essential data;
• degrade competitors' API access;
• favour affiliated lenders;
• manipulate borrower rankings;
• impose discriminatory access conditions;
• prevent lenders from exporting borrower data;
• tie AI credit scoring to payment services;
• impose exclusivity requirements.

Thus, AI coordination should not be viewed solely as a cartel problem.

It may also involve:

dominance + data + interoperability + foreclosure.

19. Alternative Data and Collective Market Power

A unique feature of AI lending is the importance of data.

Suppose one platform controls:

• payment data;
• e-commerce purchases;
• mobile-wallet transactions;
• employment information;
• behavioural data;
• repayment histories.

Competitors may become dependent upon that data to develop effective credit models.

A dominant platform could potentially use this advantage to:

• deny data access;
• impose discriminatory licensing terms;
• bundle scoring with lending;
• favour affiliated lenders;
• restrict data portability.

Competition authorities may therefore examine data access and AI model dependency alongside traditional market-share analysis.

20. Compliance Measures for Alternative Lenders

Alternative lenders should consider implementing an AI competition compliance framework.

A. Data controls

Prohibit unnecessary collection of:

• competitors' current pricing;
• future pricing plans;
• confidential strategies;
• individual competitor loan information.

B. Model governance

Document:

• model objectives;
• training data;
• feature selection;
• model updates;
• optimisation criteria.

C. Vendor due diligence

Contracts with AI vendors should specify:

• data segregation;
• confidentiality;
• no unauthorised competitor-data pooling;
• permitted model training;
• audit rights;
• access controls.

D. Independent pricing

Lenders should retain genuine authority to:

• reject recommendations;
• modify prices;
• change underwriting criteria;
• respond independently to market conditions.

E. Audit logs

Maintain records showing:

• who changed model parameters;
• why changes were made;
• what data was used;
• which recommendations were generated;
• whether recommendations were accepted or rejected.

21. Competition-Law Risk Matrix

22. Key Legal Principles Emerging From the Cases

The cases collectively support several important principles.

Principle 1 — Technology does not immunise cartel conduct

An unlawful agreement remains unlawful when implemented through software.

Topkins is particularly illustrative.

Principle 2 — Information can itself be competitively sensitive

Container Corporation demonstrates the importance of competitor pricing information.

Principle 3 — Electronic systems can facilitate concerted practices

Eturas demonstrates that competition law can operate in technologically mediated environments.

Principle 4 — Forward-looking credit information is particularly sensitive

The CJEU's recent credit-market information-exchange jurisprudence highlights the sensitivity of future credit spreads and risk variables.

Principle 5 — Common algorithms create hub-and-spoke risks

RealPage illustrates how a common algorithmic intermediary can become central to an alleged coordination theory.

Principle 6 — Parallel algorithmic outcomes do not automatically prove an agreement

Evidence concerning communications, data-sharing, common design, participation and plus factors remains important.

23. Application to Fintech and Alternative Lending

A hypothetical high-risk structure would be:

Competing lenders
↓
Share current loan-pricing and risk data
↓
Common AI vendor
↓
AI analyses competitors' confidential data
↓
Generates recommended interest rates
↓
Lenders agree to follow recommendations
↓
Reduced independent price competition

The strongest competition-law concern would arise not merely because the AI produces similar prices, but because the lenders knowingly created a system through which competitively sensitive information was exchanged and pricing decisions became coordinated.

24. Defences and Counterarguments

An alternative lender accused of AI coordination may argue that:

1. the data was publicly available;
2. information was sufficiently aggregated;
3. data was historical rather than current;
4. each lender independently selected its price;
5. the AI recommendation was non-binding;
6. lenders could freely reject recommendations;
7. there was no communication between competitors;
8. the model was trained only on the lender's own data;
9. similar prices resulted from similar economic conditions; or
10. the AI provider did not disclose competitor-specific information.

These factors can be highly significant.

Therefore, similar AI-generated lending prices should not automatically be equated with unlawful coordination.

25. Conclusion

Alternative Lending AI Coordination Risks represent a developing intersection between fintech, artificial intelligence, data governance and competition law.

The principal risks arise from:

• common AI pricing systems;
• competitor-data pooling;
• sharing future lending strategies;
• algorithmically coordinated interest rates;
• common underwriting thresholds;
• borrower allocation;
• common AI vendors;
• hub-and-spoke arrangements;
• autonomous algorithmic coordination; and
• dominant AI platforms controlling access to critical lending data.

The most important cases for building the legal framework are United States v. Topkins, Eturas, T-Mobile Netherlands, United States v. Container Corporation of America, RealPage litigation, and the emerging EU credit-information exchange jurisprudence. Together, they demonstrate that competition law focuses not on whether coordination occurred through humans, software, or AI, but on the nature of the information, the relationship between competitors, the existence of concertation or agreement, and the effect or object of the conduct.