Alternative Lending AI Coordination Risks

1. Introduction

Alternative lending refers to credit supplied through fintech lenders, peer-to-peer platforms, embedded-finance providers, marketplace lenders, BNPL providers, digital lenders and other non-traditional credit intermediaries. These businesses increasingly use artificial intelligence (AI), machine learning, alternative data and automated pricing systems for:

• borrower risk assessment;
• interest-rate determination;
• credit limits;
• loan approval or rejection;
• borrower segmentation;
• default prediction;
• collection strategies; and
• dynamic repricing.

AI can increase competition by lowering underwriting costs and enabling lenders to serve borrowers who lack conventional credit histories. The competition-law concern arises when competing lenders use the same AI system, common data pool, shared vendor, common benchmark, or algorithmic feedback mechanism in a manner that facilitates coordination.

The important distinction is between:

Independent algorithmic parallelism and algorithmically facilitated concerted conduct.

The fact that several AI systems independently produce similar lending rates does not, by itself, establish an antitrust violation. The legal risk becomes considerably greater where lenders share competitively sensitive information, deliberately use a common coordination mechanism, communicate through a technological intermediary, or knowingly delegate important competitive decisions to a common system.

Regulators have separately emphasized that AI does not remove ordinary lending-law obligations. For example, the CFPB has stated that creditors using complex or “black-box” models remain subject to specific adverse-action explanation requirements under ECOA and Regulation B.

2. How AI Coordination Can Arise in Alternative Lending

A. Common AI underwriting platform

Suppose ten competing fintech lenders purchase underwriting software from the same AI provider.

The provider receives:

• borrower income;
• default rates;
• loan acceptance rates;
• competitors' interest rates;
• rejection rates;
• credit limits;
• delinquency information; and
• conversion data.

If the system uses information from one lender to recommend pricing to another, the platform may become a coordination hub.

The central issue is not simply that lenders use the same software. It is whether the software facilitates the transmission or exploitation of non-public competitively sensitive information.

B. Algorithmic interest-rate coordination

Alternative lenders frequently compete on:

• APR;
• origination fees;
• repayment periods;
• late-payment charges;
• credit limits;
• promotional discounts.

An AI system could continuously observe market prices and recommend:

“Increase APR by 0.75% because competing lenders have increased their rates.”

If every competitor independently receives such recommendations, the market may experience rapid parallel price movements.

Parallel pricing alone is generally insufficient to establish an agreement. However, the risk increases where competitors have agreed to use a common algorithm knowing that it incorporates competitors' confidential pricing information.

3. Hub-and-Spoke AI Coordination

A particularly important theory is the hub-and-spoke model.

Traditional model

Lender A ↔ AI Vendor ↔ Lender B ↔ Lender C

The AI vendor becomes the hub.

The lenders are the spokes.

If the hub collects confidential information from competing lenders and uses it to generate coordinated recommendations, authorities may examine whether the arrangement constitutes an agreement or concerted practice.

The DOJ's RealPage litigation illustrates the contemporary application of this theory: the government alleges that competing landlords supplied non-public information to a common algorithmic pricing system and that the system generated recommendations using that pooled information. The government's theory specifically invokes a hub-and-spoke structure.

Although RealPage concerns rental pricing rather than lending, its reasoning is highly relevant to AI-based financial intermediaries.

4. Six Important Case Laws

1. United States v. David Topkins (2015)

Facts

David Topkins and co-conspirators sold posters through Amazon Marketplace. According to the DOJ, the participants agreed to fix prices and deliberately implemented their agreement through pricing algorithms.

Topkins pleaded guilty to participating in the price-fixing conspiracy.

Principle

The use of an algorithm does not immunize an otherwise conventional cartel.

The important feature was that the competitors had a human agreement and then used software to implement that agreement.

Relevance to alternative lending

Imagine competing digital lenders agree:

“We will maintain a minimum APR of 18%.”

They then program their AI systems to avoid pricing below that level.

The software is merely the mechanism implementing the underlying agreement.

Legal lesson

Human agreement + algorithmic implementation = conventional cartel liability.

This is the clearest established precedent showing that algorithmic implementation does not change the fundamental antitrust analysis.

2. Trod Ltd / GB Eye – UK CMA (2016)

The UK Competition and Markets Authority investigated sellers of posters and frames on Amazon Marketplace.

Trod and GB Eye had an agreement not to undercut each other's prices, with algorithmic pricing software used to implement the arrangement. Trod was found to have infringed UK competition law, and the CMA also pursued director-disqualification consequences.

Relevance

The case demonstrates that automated pricing can make implementation of an existing agreement:

• faster;
• more systematic;
• more difficult to monitor manually; and
• more comprehensive.

Alternative-lending analogy

Several fintech lenders might agree informally that they will not offer:

• APR below a particular level;
• zero-fee loans;
• longer repayment periods; or
• unusually high credit limits.

Their AI systems could automatically enforce that common strategy.

Principle

Automation does not transform coordinated commercial conduct into unilateral conduct.

3. Eturas UAB and Others v Lietuvos Respublikos konkurencijos taryba

CJEU, Case C-74/14 (2016)

This is one of the most important European algorithmic-coordination precedents.

Facts

Travel agencies used a common online booking system operated by Eturas. The system administrator introduced a technical restriction limiting the discounts that agencies could offer.

The issue was whether participating businesses could be treated as having participated in an anti-competitive concerted practice merely because the restriction was introduced through the common electronic system.

The CJEU examined knowledge, participation and the circumstances from which awareness of the anti-competitive mechanism could be established.

Relevance to AI lending

Consider a common lending platform used by competing lenders.

The platform introduces:

“Recommended minimum APR = 16%.”

If participating lenders know that the common platform is imposing or facilitating this pricing restriction and continue using it, the legal analysis becomes much more serious than simple independent use of similar technology.

Principle

Electronic or automated communication can constitute the mechanism through which competitors become aware of and participate in coordinated conduct.

4. United States v. Apple Inc. — E-books (2d Cir. 2015)

Although not an AI case, Apple is highly relevant to intermediary-facilitated coordination.

The Second Circuit upheld findings concerning Apple's role in orchestrating coordination among major publishers concerning e-book pricing. The court concluded that the evidence supported the finding of a horizontal conspiracy facilitated through Apple's contractual and intermediary relationships.

Why it matters for alternative lending

An AI platform can potentially perform a similar intermediary function.

For example:

Lender A → AI platform ← Lender B

If the platform:

1. collects information from competing lenders;
2. communicates market information;
3. encourages lenders to adopt a common pricing strategy; and
4. knows that the lenders are competing with each other,

the platform may potentially become more than an ordinary software supplier.

Principle

A vertically positioned intermediary can potentially facilitate horizontal coordination among otherwise competing businesses.

5. United States v. RealPage, Inc. (2024–2026 proceedings)

This is currently one of the most important modern algorithmic-coordination proceedings.

The DOJ alleges that RealPage's system collected competitively sensitive information from competing landlords and used that information to generate pricing recommendations. The government characterizes the alleged structure as having features of a hub-and-spoke conspiracy.

Relevance to alternative lending

The same architecture could theoretically exist in lending:

Competing lenders

↓

Common AI underwriting/pricing vendor

↓

AI model trained on confidential lender data

↓

Recommended APR / credit limit / fees

The critical question would be whether lenders knowingly participate in a system that facilitates coordination rather than merely using a neutral analytical product.

Important qualification

RealPage is not a final judicial determination that every common algorithm constitutes an antitrust violation. The allegations and procedural developments must be distinguished from a final merits judgment.

Nevertheless, the case illustrates the enforcement direction concerning common algorithmic intermediaries.

6. Interstate Circuit, Inc. v. United States

306 U.S. 208 (1939)

This is a foundational U.S. hub-and-spoke case.

Facts

A group of distributors received demands from a common intermediary concerning minimum resale conditions. The Supreme Court considered whether the circumstances supported an inference of concerted action even though direct communications between all participants were not necessary.

Relevance to AI lending

An AI vendor could theoretically play the role of a modern technological intermediary.

For example:

Lender A gives confidential information to Vendor X.
Lender B gives confidential information to Vendor X.
Vendor X communicates recommendations based on the combined information.

The absence of direct lender-to-lender communication would not necessarily end the inquiry.

Principle

Competition law can examine indirect coordination through an intermediary, particularly where the surrounding circumstances demonstrate participation in a common scheme.

5. Why Alternative Lending Is Particularly Vulnerable

AI creates several distinctive competition risks in lending.

A. Common data pools

A fintech may possess information concerning:

• borrower default probability;
• average APR;
• repayment behaviour;
• loan acceptance;
• rejected applications;
• customer switching;
• credit utilization;
• geographic risk;
• borrower elasticity.

If competing lenders contribute this information to a common AI model, the model could potentially reduce the competitive uncertainty that normally exists between lenders.

B. Common pricing engines

A single vendor may provide AI pricing infrastructure to many lenders.

The danger increases if the vendor:

• sees each lender's prices;
• knows each lender's intended future prices;
• receives confidential transaction data;
• recommends future prices;
• uses one lender's data to improve recommendations to another; or
• encourages adoption of the same recommendations.

6. Tacit Coordination Without Express Agreement

The more difficult problem is autonomous AI coordination.

Suppose five lenders independently instruct reinforcement-learning systems:

“Maximize long-term lending profits.”

The algorithms observe market prices and repeatedly modify APRs.

Over time they could theoretically learn:

“If I raise my rate, competitors follow.”

The algorithms may therefore converge on supra-competitive outcomes without any human agreeing:

“Let us fix prices.”

This creates an important distinction:

Type 1 — Express coordination

Humans agree to coordinate → AI implements it.

Highest traditional antitrust risk.

Type 2 — Facilitated coordination

Competitors use a common AI intermediary containing confidential competitor information.

Significant potential risk depending on knowledge, participation and market effects.

Type 3 — Autonomous algorithmic convergence

Independent AI systems independently learn similar pricing behaviour.

More difficult legal question because proof of agreement or concerted action may be absent.

7. Alternative Data and Competitive Coordination

Alternative lenders increasingly use:

• transaction data;
• mobile-device information;
• bank-account data;
• e-commerce history;
• employment information;
• cash-flow information;
• telecommunications data;
• behavioural data.

The CFPB has recognized that alternative data and machine learning can affect both credit access and fair-lending outcomes.

From a competition perspective, a major concern arises where a dominant data intermediary controls information necessary for competing lenders to develop comparable models.

This creates two separate questions:

Data-exclusion question

Can rival lenders obtain sufficiently comparable data?

Coordination question

Are competing lenders sharing sensitive data through the same AI infrastructure?

These should not be conflated.

8. AI Vendor as a Potential Competition Bottleneck

An AI vendor may occupy a strategically important position where many lenders depend upon it for:

• credit scoring;
• fraud detection;
• underwriting;
• pricing;
• collections;
• borrower segmentation.

If switching costs are high, the vendor may become an important intermediary between competing lenders.

Potential competition concerns include:

1. discriminatory access to model functionality;
2. preferential treatment of large lenders;
3. use of one lender's data for another lender;
4. restrictions on interoperability;
5. exclusionary contracts;
6. tying underwriting to other financial services;
7. exclusive data arrangements;
8. common pricing recommendations; and
9. facilitation of coordinated conduct.

9. Competition Law Versus Consumer Protection Law

AI lending can simultaneously create two different categories of legal issues.

The CFPB has specifically stated that creditors cannot avoid ECOA obligations simply because an AI model is complex or opaque.

Thus, an alternative lender could potentially face both competition-law and consumer-finance scrutiny from the same AI system, but the legal tests are different.

10. India — Competition Act, 2002 Perspective

For an Indian alternative-lending ecosystem, the central competition provisions would principally involve Section 3 concerning anti-competitive agreements and Section 4 concerning abuse of dominant position.

Section 3 concerns

Potential issues include:

• coordinated lending rates;
• coordinated fees;
• market allocation;
• sharing competitively sensitive lending information;
• common algorithmic pricing arrangements;
• restrictions imposed through fintech intermediaries.

The technological form of the arrangement should not determine whether substantive competition law applies.

Section 4 concerns

A dominant digital lending or AI infrastructure provider could potentially raise concerns concerning:

• discriminatory access;
• denial of access;
• unfair conditions;
• leveraging;
• tying;
• exclusionary interoperability restrictions; and
• exploitation of dependency.

The particularly difficult issue is whether an AI provider is merely a technology supplier or has become an important competitive intermediary with sufficient market power.

11. Key Evidentiary Questions

Competition authorities investigating AI coordination would likely need to examine evidence such as:

Internal communications

• emails;
• WhatsApp/Slack messages;
• board papers;
• vendor contracts;
• pricing committee documents.

Algorithmic records

• model versions;
• training datasets;
• prompts;
• system instructions;
• pricing recommendations;
• model logs;
• API calls;
• override records.

Data architecture

Authorities may ask:

Whose data entered the model?

Who could access it?

Was competitor data identifiable?

Was it aggregated?

Was it sufficiently historical to reduce competitive sensitivity?

Did one lender's information affect another lender's recommendation?

Behavioural evidence

Authorities may compare:

• rates before and after algorithm adoption;
• pricing dispersion;
• lender responses to competitors;
• frequency of algorithm acceptance;
• deviations from recommendations.

12. Compliance Framework for Alternative Lenders

A robust compliance program should include:

1. Competitor-data firewall

Do not allow confidential lender A information to influence lender B's pricing recommendations unless legally justified and appropriately aggregated/anonymized.

2. Vendor due diligence

Contracts with AI providers should specify:

• data ownership;
• data segregation;
• permissible data uses;
• model-training restrictions;
• confidentiality;
• audit rights;
• deletion obligations.

3. Independent pricing decisions

Lenders should retain meaningful control over:

• APR;
• fees;
• credit limits;
• loan terms.

4. Algorithmic audit trails

Maintain records showing:

• data inputs;
• model versions;
• recommendations;
• human overrides;
• pricing decisions.

5. Competition-law review

Competition counsel should review AI systems that:

• process competitor data;
• recommend prices;
• coordinate market behaviour;
• benchmark competitors;
• facilitate information exchanges.

6. No “competitor-aware” optimization without review

A system specifically designed to maximize returns by responding to competitors' confidential future pricing should receive heightened antitrust scrutiny.

13. Important Distinction: Similar Prices ≠ Automatically Collusion

This is crucial for examination purposes.

Suppose ten alternative lenders independently use similar public data.

Their AI systems calculate:

Lender A — 14.9%
Lender B — 15.0%
Lender C — 14.8%

The similarity of prices does not itself establish an agreement.

The analysis changes if evidence shows:

All three lenders supplied confidential pricing data to the same intermediary, knew that their competitors were doing so, and knowingly adopted the intermediary's recommendations generated from the pooled information.

The second scenario raises substantially stronger coordination concerns.

14. Emerging Autonomous-AI Problem

The most difficult future question is:

Can competition law address coordination created by AI systems when no human directly communicated with a competitor?

Traditional cartel law generally searches for some form of:

• agreement;
• concerted practice;
• communication;
• conscious commitment;
• participation; or
• facilitating conduct.

Autonomous AI creates a possible gap:

Human Lender A → AI Agent A

Human Lender B → AI Agent B

↓

Independent AI learning

↓

Common supra-competitive outcome

The economic outcome may resemble coordination even when the conventional evidence of a human agreement is absent.

That distinction is one reason current algorithmic-collusion debates focus heavily on the design, deployment, information environment and governance of algorithms, rather than simply on whether prices moved together.

15. Case-Law Principles at a Glance

16. Conclusion

Alternative Lending AI Coordination Risks arise at the intersection of antitrust law, algorithmic decision-making, alternative data and fintech infrastructure.

The most important legal distinction is between:

AI independently observing the market
and
AI facilitating coordination between competitors.

The greatest risks arise where competing lenders:

1. use a common AI intermediary;
2. contribute non-public competitively sensitive information;
3. receive recommendations based on competitors' information;
4. knowingly adopt common pricing strategies;
5. communicate indirectly through the algorithm;
6. delegate competitive decisions to a common system; or
7. use AI to stabilize prices or other competitive variables.

Topkins and Trod demonstrate that algorithms do not shield conventional price fixing. Eturas demonstrates the relevance of a common electronic platform. Apple and Interstate Circuit provide broader intermediary/hub-and-spoke principles. RealPage represents the particularly important modern application of those concepts to algorithmic pricing, although its allegations should not be treated as a final adjudication of every algorithmic-pricing model.