Competition Law and Machine-Driven Credit Allocation and Competition

Jurisdictional approach: Indian competition law, supplemented by comparative authorities. There is still relatively little reported Indian competition case law directly deciding AI-driven credit allocation. Therefore, some of the authorities below are direct digital-finance/competition cases, while others are analogical competition precedents dealing with platform power, information, algorithms, market access and technological control.

India's regulatory framework is particularly relevant because the RBI has recognised that digital lenders increasingly use proprietary algorithms for credit underwriting and has recommended that underwriting algorithms be auditable for minimum standards and potential discriminatory factors. The RBI has also identified concentration, market-power and systemic-risk concerns arising from technology-driven digital lending. (System Health) The CCI released its Market Study Report on Artificial Intelligence and Competition in October 2025, showing that AI-related competitive issues are now an explicit part of Indian competition-policy analysis. (Competition Commission of India)

1. Meaning of Machine-Driven Credit Allocation

Machine-driven credit allocation means using algorithms, artificial intelligence, machine learning, automated scoring systems and data analytics to decide:

who receives credit;

how much credit is offered;

interest rates;

credit limits;

repayment conditions;

loan eligibility;

risk classification;

customer targeting;

loan approval or rejection.

A simplified structure is:

Applicant data → AI/ML model → Credit score → Risk assessment → Loan decision → Pricing

The competition-law question is different from the ordinary consumer-protection question.

The central competition question is:

Can control over credit-allocation technology, data, algorithms or digital lending infrastructure create, strengthen or abuse market power?

2. Machine-Driven Credit Markets

Modern credit markets can involve:

banks;

NBFCs;

fintech companies;

digital lending platforms;

payment platforms;

credit-information companies;

e-commerce platforms;

BigTech firms;

embedded-finance providers.

Machines can connect all of these participants.

For example:

E-commerce platform → customer data → AI credit score → lender → instant loan

This creates a potentially powerful ecosystem.

3. Competition Law vs Financial Regulation

Credit allocation is heavily regulated.

Competition law therefore operates alongside:

RBI regulation;

banking regulation;

NBFC regulation;

consumer-protection law;

data-protection requirements;

financial-sector regulation.

The CCI's role is to protect competition, while the RBI regulates financial stability and regulated lending entities.

The RBI's digital-lending framework expressly states that outsourcing to a Lending Service Provider does not remove the regulated entity's regulatory obligations. (System Health)

4. Why Machine-Driven Credit Creates Competition Issues

Traditional lending competition was based largely on:

branch networks;

capital;

interest rates;

customer relationships.

Machine lending adds:

data;

AI;

computing;

predictive models;

digital distribution;

network effects;

automated underwriting.

Consequently, a company may gain competitive power through data and technology, even without controlling the largest amount of financial capital.

5. Data as a Competitive Advantage

Credit algorithms may use:

transaction history;

repayment history;

banking information;

income information;

purchase behaviour;

payment behaviour;

business activity;

digital-platform activity.

The RBI's digital-lending working-group report specifically described digital-lending applications as potentially using information such as credit scores, historical banking information and mobile-recharge information for proprietary scoring and underwriting. (System Health)

If a company has access to a uniquely valuable dataset, competitors may find it difficult to reproduce the same credit model.

6. Data Feedback Loops

A machine-credit platform can create:

More customers → more data → better model → better risk prediction → more customers → more data.

This is a data feedback loop.

Over time, it may increase the competitive advantage of the incumbent.

This does not automatically establish dominance, but it can contribute to market power.

7. Network Effects

Credit platforms can also experience network effects.

More borrowers may attract:

more lenders;

more merchants;

more transaction data;

more payment activity.

More lenders may produce:

better product variety;

greater transaction volume;

more data.

This can create:

Borrowers ↔ Platform ↔ Lenders

The platform may therefore become increasingly important to both sides.

8. Machine-Driven Credit Scoring

AI credit scoring generally involves:

collecting data;

cleaning data;

identifying variables;

training a model;

calculating risk;

assigning a score;

determining eligibility;

determining price;

monitoring repayment;

updating the model.

The model may continuously learn from new borrowers.

9. Competition Concern: Access to Data

Suppose Platform A controls:

millions of customer transactions;

payment history;

shopping history;

repayment data.

Competitor B cannot access equivalent data.

Platform A may therefore have a significant informational advantage.

Competition law may become relevant where control over such data is used in an exclusionary manner by a dominant enterprise.

10. Machine-Driven Credit and Section 3

Section 3 of the Competition Act concerns anti-competitive agreements.

Potential problems could occur if competing lenders use a common AI system to:

coordinate interest rates;

coordinate credit limits;

exchange future pricing information;

allocate borrowers;

restrict lending;

coordinate risk policies.

For example:

Five competing lenders provide future interest-rate information to a common algorithm, and the system recommends identical rates.

The technology does not itself establish an infringement, but the underlying information exchange and coordinated conduct could become relevant under Section 3.

11. Machine-Driven Credit and Section 4

Section 4 becomes particularly important where a technology company or financial platform has a dominant position.

Potential abusive conduct could include:

discriminatory access;

denial of market access;

tying;

self-preferencing;

exclusionary data practices;

discriminatory ranking;

predatory strategies.

12. Self-Preferencing in Credit Markets

Imagine an e-commerce platform owns a lending service.

It operates:

Marketplace + Payment System + AI Credit Score + Lending Service

The platform's algorithm could potentially recommend its own loans before competing lenders.

This creates a potential self-preferencing issue if the necessary dominance and abuse requirements are established.

The CCI's digital-platform cases demonstrate why control over platform infrastructure and preferential treatment can have competition significance. (Competition Commission of India)

13. Credit Marketplace Gatekeepers

A digital platform may become a gateway between:

Borrower → Credit marketplace → Lender

If a platform becomes sufficiently important, competing lenders may become dependent upon it for customer acquisition.

The platform could potentially control:

customer visibility;

rankings;

access to data;

lead generation;

loan recommendations.

This creates possible gatekeeper power.

14. Algorithmic Ranking of Loans

Suppose a platform displays:

its own loan;

affiliated lender;

independent lender;

another independent lender.

If ranking is determined by legitimate factors such as price or customer suitability, it may be pro-competitive.

But if a dominant platform systematically favours its affiliated lending service without adequate competitive justification, the conduct may warrant examination.

15. Credit Allocation and Discrimination

Machine learning can create different outcomes for different applicants.

For example:

Applicant A receives ₹5 lakh;

Applicant B receives ₹1 lakh;

Applicant C receives no loan.

Differentiation can be economically justified.

However, competition concerns can arise if algorithmic discrimination is used strategically to:

exclude competitors;

foreclose alternative lenders;

discriminate against customers who use rival services;

condition access to one service on another.

16. Personalised Pricing

AI may determine individualised interest rates.

For example:

Customer 1 → 10%
Customer 2 → 13%
Customer 3 → 18%

Individualised pricing is not automatically an antitrust violation.

But a dominant platform using personalised pricing to exploit or exclude customers may create Section 4 concerns depending on the facts.

17. Predatory Credit Strategies

A dominant fintech could theoretically provide unusually cheap loans to:

eliminate a rival lender;

capture customers;

increase market share;

make entry commercially unattractive.

After competitors exit, prices could potentially rise.

Such conduct would need to satisfy the legal requirements applicable to predatory pricing or other forms of abuse.

18. Credit Allocation as a Market-Access Tool

Credit can determine whether businesses survive.

For example:

Small manufacturer → working-capital loan → production → market participation

If a dominant financial platform controls access to necessary financing and deliberately excludes rival businesses from credit, the competition implications can become significant.

19. Credit and SME Competition

Small businesses may depend heavily upon:

working capital;

invoice financing;

merchant loans;

supply-chain financing.

AI can make credit allocation faster.

But if one platform controls the relevant business data and financing infrastructure, competitors may face barriers to entry.

20. Embedded Finance

Embedded finance integrates credit into another service.

Examples:

e-commerce + business loan;

payment app + personal loan;

accounting software + working-capital finance;

ride-hailing platform + driver loan.

This can produce strong ecosystem effects.

21. Cross-Market Leveraging

A platform might possess market power in:

Payment services

and use that position to expand into:

Credit

or possess power in:

E-commerce

and leverage it into:

Merchant lending.

Competition law may examine whether power in one market is being used to restrict competition in another.

22. Tying Credit to Other Services

Potential example:

"You will receive the loan only if you use our payment service."

Or:

"Merchant credit is available only to sellers who use our payment gateway."

Such conduct may require examination under the relevant tying/bundling provisions where the enterprise has the requisite market position and the statutory conditions are met.

23. Exclusivity

A platform could require merchants to:

use its lender;

use its payment system;

avoid competing credit providers.

Exclusive arrangements may produce efficiencies but may also foreclose competing lenders depending on:

duration;

market coverage;

market power;

alternatives;

foreclosure effects.

24. Machine-Driven Credit and Switching Costs

Borrowers may become locked into a platform because it possesses:

credit history;

transaction history;

customised pricing;

repayment records;

financial data.

Moving to another lender may therefore become costly.

Data portability and interoperability can potentially reduce these switching costs.

25. Credit-Scoring Infrastructure as a Bottleneck

Suppose one platform becomes the dominant provider of an AI credit score.

Lenders may increasingly depend upon it.

This could create a bottleneck in:

credit assessment;

customer identification;

risk evaluation.

The competition question becomes whether competitors have reasonable alternatives.

26. Third-Party Credit Algorithms

Many lenders may purchase underwriting technology from the same vendor.

For example:

Lender A → AI Provider ← Lender B

The provider may receive data from multiple competing lenders.

This can create risks involving:

confidential information;

algorithmic coordination;

common pricing models;

standardised credit decisions.

27. Algorithmic Coordination of Interest Rates

Suppose competing lenders use one algorithm that:

observes competitor rates;

predicts reactions;

adjusts rates automatically.

The algorithm could potentially facilitate coordination.

This resembles broader algorithmic-collusion problems.

The crucial distinction remains:

independent algorithmic pricing versus deliberate or legally cognizable coordination.

28. Machine Learning and Pro-Cyclicality

The RBI has highlighted another important risk: widespread reliance on automated underwriting and similar optimisation algorithms can potentially amplify systemic and pro-cyclical risks. (System Health)

For competition law, this matters because many lenders adopting similar algorithms can create:

simultaneous tightening of credit;

simultaneous expansion;

similar risk classifications;

concentrated lending patterns.

This does not automatically establish an antitrust violation, but it can affect market structure and competitive resilience.

29. Credit Algorithm and Entry Barriers

New lenders may need:

large datasets;

AI infrastructure;

historical repayment information;

customer networks.

An incumbent with all these resources may have substantial advantages.

Thus, machine-driven credit can increase endogenous entry barriers.

30. BigTech Entry into Credit Markets

A large technology company may already possess:

payment data;

search data;

e-commerce data;

advertising data;

identity infrastructure;

cloud infrastructure.

Entering credit markets can allow it to combine these assets.

This raises the possibility of cross-market leveraging.

The RBI has expressly recognised potential concentration and competition concerns if BigTech entities enter direct digital lending markets. (System Health)

31. Data Combination

Consider:

E-commerce data + payment data + search data + lending data

A single enterprise could construct an exceptionally detailed customer profile.

The competitive concern is not simply privacy.

It may also concern whether rivals can compete without equivalent data.

32. Data Exclusivity

A dominant platform might restrict competitors from accessing relevant data.

Possible effects:

higher entry barriers;

weaker innovation;

reduced lender choice;

increased dependence on the platform.

Competition authorities may therefore examine whether data access forms part of an exclusionary strategy.

33. Machine Credit and Financial Inclusion

AI credit allocation can also be strongly pro-competitive.

It may:

lower underwriting costs;

reach underserved borrowers;

reduce processing time;

enable small-ticket loans;

reduce information asymmetry;

increase lender competition.

Competition policy must therefore avoid treating AI credit as inherently harmful.

34. Efficiency Defence

A lender may legitimately argue that its AI system:

reduces default risk;

reduces transaction costs;

improves fraud detection;

lowers interest rates;

increases credit availability.

Such efficiencies must be considered within the applicable competition-law framework.

35. Case Law 1 – Satyen Narendra Bajaj v. PayU Payments Pvt. Ltd., CCI Case No. 23/2019

The CCI examined a competition complaint involving PayU in the digital-payments ecosystem. (Competition Commission of India)

Importance

It demonstrates the CCI's willingness to examine competition issues involving digital financial platforms.

Relevance to machine-driven credit

Modern credit platforms increasingly combine:

payments;

transaction data;

digital identity;

financial services;

lending.

Therefore, competition analysis may need to consider the broader digital-finance ecosystem rather than viewing credit in isolation.

36. Case Law 2 – Google Android / Umar Javeed v. Google

The CCI's Android proceedings concerned Google's position across interconnected digital markets and the relationship between Android, app distribution and other services. The CCI imposed a penalty in 2022 for anti-competitive practices concerning Android mobile devices. (Competition Commission of India)

Principle

Control over an important digital platform can influence competition in adjacent markets.

Relevance

A financial platform that controls:

payments + customer interface + data + credit

could similarly possess ecosystem advantages that affect competing lenders.

37. Case Law 3 – Google Play Store Proceedings

The CCI found competition concerns relating to Google's Play Store policies and imposed a monetary penalty in 2022. (Competition Commission of India)

The CCI's findings concerning preferential treatment and access to payment-related functionality are particularly relevant to digital financial ecosystems. (Competition Commission of India)

Relevance

The case illustrates how control over a digital distribution layer can affect competing financial or payment services.

The same structural principle can apply to:

Credit platform → loan ranking → lender access → customer acquisition.

38. Case Law 4 – CCI v. SAIL, (2010) 10 SCC 744

This Supreme Court decision remains a foundational Indian competition-law authority.

Relevance

Machine-driven credit investigations may involve extensive:

digital evidence;

data analysis;

algorithmic records;

economic evidence.

The CCI's statutory investigation framework therefore provides the foundation for examining emerging technology markets.

39. Case Law 5 – Excel Crop Care Ltd. v. CCI, (2017) 8 SCC 47

The Supreme Court examined anti-competitive conduct and penalty principles.

Relevance

The important lesson for AI lending is that the technological mechanism does not determine the legality of conduct.

An enterprise cannot avoid competition-law scrutiny merely because:

"The algorithm made the decision."

The underlying enterprise conduct and economic consequences remain important.

40. Case Law 6 – CCI v. Bharti Airtel Ltd., (2019) 2 SCC 521

The Supreme Court addressed the relationship between competition law and sector-specific regulation.

Relevance

Machine-driven credit exists in a highly regulated financial environment.

Consequently, cases involving:

RBI regulation;

banking;

NBFCs;

payment systems;

digital lending

may involve both sectoral regulation and competition law.

Bharti Airtel is therefore important for understanding how these regulatory spheres can interact.

41. Case Law 7 – Ohio v. American Express Co., 585 U.S. 529 (2018)

The US Supreme Court dealt with competition in a two-sided transaction platform.

Principle

Some platforms must be analysed by considering the interaction between different sides of the platform.

Relevance to machine credit

A credit platform may connect:

Borrowers ↔ Platform ↔ Lenders

or:

Merchants ↔ Platform ↔ Financial institutions

Competition effects may therefore need to be considered across the platform rather than on only one side.

42. Case Law 8 – United States v. Microsoft Corp., 253 F.3d 34 (D.C. Cir. 2001)

Microsoft is an important technological-platform precedent.

Principle

A company with substantial platform power may potentially use that position to affect competition in adjacent technological markets.

Relevance

A financial technology ecosystem could similarly connect:

Payments → Data → Credit scoring → Lending → Insurance

Control of the first layer could potentially be leveraged into the credit market.

43. Case Law 9 – United States v. Apple Inc., 791 F.3d 290 (2d Cir. 2015)

The Apple e-books litigation demonstrates how technology platforms and contractual arrangements can facilitate coordination among market participants.

Relevance

A common credit algorithm or digital lending platform may similarly become an intermediary through which competing lenders interact.

The factual and legal requirements of the respective competition laws remain different, so the analogy should not be treated as a direct holding on AI credit scoring.

44. Case Law 10 – Eturas UAB and Others v. Lietuvos Respublikos konkurencijos taryba, Case C-74/14

The CJEU considered a common computerized system used by travel agencies, including an automated restriction on discounts.

Principle

A computerized system can be relevant evidence when assessing concerted practices.

Relevance to credit

If competing lenders use a common automated system to coordinate:

interest rates;

fees;

credit limits;

the system itself may become important evidence.

45. Direct vs Analogical Authorities

46. Machine Credit and Market Definition

A credit ecosystem may contain several relevant markets:

Consumer credit

Personal loans and consumer financing.

Merchant credit

Financing for businesses selling through platforms.

Digital lending

Loans originated/distributed through digital platforms.

Credit-scoring services

Technology providing credit-risk assessment.

Lending infrastructure

Technology used by lenders for underwriting and servicing.

Embedded finance

Credit integrated into non-financial platforms.

The relevant market must be determined according to competitive substitutability and market conditions rather than assuming that all financial services form one market.

47. Credit-Scoring Market Power

Suppose a company controls a highly important credit-scoring system.

It might potentially:

determine which lenders receive customers;

influence loan pricing;

rank lenders;

restrict data access.

The competition question becomes whether the scoring platform possesses substantial market power and whether its conduct restricts competition.

48. Refusal to Share Data

A dominant platform might refuse to provide competitors access to important data.

A refusal is not automatically unlawful.

Competition analysis must consider factors such as:

market power;

indispensability;

alternatives;

justification;

competitive effects;

impact on innovation.

49. Credit Platform Self-Preferencing

Consider:

Platform A owns marketplace + payment system + lender.

Its AI recommends:

"Platform A Loan – Recommended"

while competing lenders receive lower visibility.

Potential issues include:

self-preferencing;

discriminatory ranking;

leveraging;

denial of market access.

The relevant legal elements must still be independently established.

50. Machine Credit and Consumer Choice

Competition is not only about the number of lenders.

It can also concern:

number of offers displayed;

transparency;

switching;

ability to compare loans;

lender visibility;

interoperability.

An algorithm that controls what consumers see can therefore influence competitive conditions.

51. Explainability and Competition

The RBI's digital-lending recommendations emphasise that algorithmic underwriting should be auditable and that potential discrimination factors affecting credit availability and pricing should be identifiable. (Reserve Bank of India)

From a competition perspective, explainability can also help determine:

whether a platform favours affiliated lenders;

whether competitors are excluded;

whether ranking is discriminatory;

whether data is being used competitively.

52. Algorithmic Audits

A credit platform may use:

model validation;

independent testing;

fairness testing;

competition testing;

data audits.

An algorithm audit can identify whether:

competitor data is being used;

affiliated lenders receive preferential treatment;

customers are systematically steered;

pricing rules produce exclusionary effects.

53. Machine Credit and Merger Control

Important acquisitions may include:

fintech + bank;

credit-scoring company + lender;

payment platform + lending platform;

e-commerce platform + fintech;

AI company + credit-information business.

Competition authorities should consider whether the transaction combines:

customer access + financial data + credit scoring + lending capacity.

54. Killer Acquisitions in FinTech

A large financial platform might acquire a small AI-credit start-up.

The start-up may have:

little revenue;

few customers;

valuable technology;

innovative credit-scoring models.

The acquisition could nevertheless matter for future competition.

55. Conglomerate Effects

Suppose one company controls:

Search + Payments + E-commerce + Cloud + AI + Credit

The company could possess several complementary advantages.

Potential concerns include:

cross-subsidisation;

data combination;

tying;

self-preferencing;

foreclosure;

customer lock-in.

However, having multiple businesses is not itself unlawful.

56. Machine-Driven Credit and Small Lenders

AI can actually increase competition by reducing entry costs.

A small fintech may use:

cloud infrastructure;

open-source AI;

alternative data;

automated underwriting.

This can allow it to compete with large banks.

Therefore, competition policy should preserve access to technological infrastructure while addressing genuine exclusion.

57. Alternative Data

Alternative data can include:

transaction patterns;

merchant behaviour;

digital payments;

business cash flows.

This may improve credit access.

But if one platform has exclusive access to valuable alternative data, competitors may face an informational disadvantage.

58. Algorithmic Credit Cartels

A hypothetical cartel might operate as follows:

Bank A + Bank B + Fintech C

↓

Provide future pricing information to common AI

↓

AI recommends similar interest rates

↓

Banks automatically follow

↓

Price competition decreases.

This would require careful legal analysis of the actual communications, arrangements and enterprise conduct.

Similar algorithmic outcomes alone should not automatically be treated as proof of cartelisation.

59. Regulatory Overlap

A machine-credit case may involve:

CCI

Competition.

RBI

Banking and digital lending regulation.

Data-protection authorities/framework

Personal-data governance.

Consumer authorities

Unfair consumer practices.

Courts

Judicial review and remedies.

This makes institutional coordination important.

60. Future Competition-Law Challenges

Major future questions include:

Who is responsible for an autonomous credit decision?

Can AI-based credit scoring create market dominance?

When does data become an essential competitive input?

Can a common underwriting algorithm facilitate coordination?

Can a dominant platform favour its own lender?

How should credit-ranking algorithms be investigated?

How should competition authorities obtain proprietary model information?

How should algorithmic discrimination be distinguished from legitimate risk pricing?

How should fintech acquisitions be assessed?

How should competition law respond to AI-driven financial ecosystems?

61. Recommended Competition-Compliance Framework

Financial enterprises using machine-driven credit systems should consider:

1. Data governance

Restrict inappropriate use of competitor information.

2. Algorithm documentation

Maintain records explaining important credit decisions.

3. Competition review

Test whether algorithms could facilitate coordination or exclusion.

4. Independent pricing

Avoid mechanisms that automatically coordinate competitor pricing.

5. Platform neutrality

Ensure competing lenders receive fair access where required by law.

6. Auditability

Maintain technical records and logs.

7. Third-party provider controls

Review common algorithms supplied to competing lenders.

8. Merger review

Assess acquisitions involving credit-data and AI assets.

9. Human oversight

Ensure significant automated systems remain subject to appropriate governance.

10. Regulatory coordination

Coordinate competition compliance with applicable RBI requirements.

62. Key Distinction

The following proposition is incorrect:

"AI credit scoring is anti-competitive."

The correct approach is:

AI credit scoring may increase efficiency and competition, but it can also create competition concerns where control over data, algorithms, platforms or financial infrastructure is used to coordinate or exclude competitors.

63. Key Legal Principles

Machine-driven credit allocation is not inherently anti-competitive.

Data can become an important source of competitive advantage.

AI underwriting can lower transaction costs and increase financial inclusion.

Common algorithms can create coordination risks among competing lenders.

A dominant digital platform may potentially leverage power from payments or commerce into lending.

Self-preferencing can become relevant where a platform favours its affiliated lender.

Credit-ranking systems can influence market access.

Interoperability and data portability can affect contestability.

FinTech acquisitions may raise innovation and future-competition concerns.

Algorithmic outcomes alone do not automatically establish an unlawful agreement.

Technical evidence and algorithmic audit trails will increasingly matter in CCI investigations.

Competition law must operate alongside RBI's financial-sector regulation.

Quick Revision

Definition

Machine-driven credit allocation means the use of AI, algorithms, machine learning, automated scoring and data analytics to determine credit eligibility, amount, pricing and access.

Main competition chain

Data

↓

AI credit scoring

↓

Customer classification

↓

Automated loan allocation

↓

Platform power

↓

Potential market concentration

↓

Possible exclusion / self-preferencing / coordination

Important authorities

Satyen Narendra Bajaj v. PayU Payments, CCI Case No. 23/2019.

Umar Javeed v. Google / Google Android, CCI Case No. 39/2018.

Google Play Store proceedings, CCI Cases 07/2020, 14/2021 and 35/2021.

CCI v. SAIL, (2010) 10 SCC 744.

Excel Crop Care Ltd. v. CCI, (2017) 8 SCC 47.

CCI v. Bharti Airtel Ltd., (2019) 2 SCC 521.

Ohio v. American Express, 585 U.S. 529 (2018).

United States v. Microsoft, 253 F.3d 34 (D.C. Cir. 2001).

United States v. Apple, 791 F.3d 290 (2d Cir. 2015).

Eturas, C-74/14.

Conclusion

Machine-driven credit allocation is transforming competition in financial markets. AI can make underwriting faster, reduce information asymmetry, lower lending costs and expand access to credit. At the same time, control over financial data, credit-scoring algorithms, digital distribution, payment infrastructure and customer interfaces can create new forms of market power.

For Indian competition law, the central issues will increasingly involve Section 3 coordination, Section 4 abuse of dominance, digital-platform leveraging, data advantages, self-preferencing, algorithmic pricing, interoperability and fintech combinations. The RBI's own work recognises both the efficiency benefits of technology-driven lending and the risks associated with concentration, market power and opaque automated underwriting. (System Health)

The long-term objective should therefore be to preserve competitive access to data and financial infrastructure while allowing legitimate AI innovation to improve credit allocation and financial inclusion.