Competition Law and Machine-Directed Market Foreclosure

1. Introduction

Machine-directed market foreclosure refers to situations where AI systems, algorithms, automated decision-making tools, or machine-controlled commercial infrastructure are used in a way that restricts competitors' access to customers, suppliers, data, distribution channels, platforms, infrastructure, or other commercially important inputs.

The important competition-law question is not simply whether a machine made the decision. The question is:

Does the use or control of automated technology substantially restrict effective competition or reinforce market power by excluding actual or potential competitors?

Machine-directed foreclosure may arise in digital platforms, AI marketplaces, cloud computing, advertising, e-commerce, payment systems, logistics, labour platforms, autonomous procurement, and algorithmic distribution.

2. Meaning of Market Foreclosure

Market foreclosure occurs when competitors are prevented, weakened, or disadvantaged from obtaining meaningful access to an important market, input, customer group, distribution channel, technology, or infrastructure.

Traditional foreclosure could be created through:

exclusive contracts;

refusal to supply;

tying;

discriminatory pricing;

exclusive distribution;

loyalty rebates;

vertical integration;

control of essential infrastructure.

Machine-directed foreclosure adds another layer:

AI automatically ranks products;

algorithms determine access to customers;

automated systems impose exclusivity;

algorithms discriminate against competing suppliers;

platform algorithms favour the platform's own products;

AI systems prevent multi-homing;

automated pricing makes switching commercially unattractive;

access to data or computing resources is controlled algorithmically.

3. Basic Formula

A useful competition-law formula is:

Machine-Directed Foreclosure = Market Power + Automated Control + Strategic Restriction + Competitor Disadvantage + Foreclosure Effect

The existence of an algorithm alone is not enough.

There should generally be an identifiable connection between the conduct and a restriction of competition.

4. Why Machine-Directed Foreclosure Is Important

Machines can operate at a scale and speed that traditional contractual systems cannot.

An ordinary manager might exclude ten competitors.

An algorithm can potentially:

rank millions of products;

determine access for thousands of businesses;

change prices simultaneously;

allocate customers automatically;

restrict access to a platform;

determine search visibility;

control advertising placement;

monitor competitors continuously.

Therefore, technological foreclosure can be:

1. Continuous

The system can operate 24 hours a day.

2. Scalable

A single algorithm can affect millions of transactions.

3. Dynamic

AI can modify its behaviour according to market conditions.

4. Difficult to detect

The exclusion may be embedded in technical rules rather than an express contract.

5. Self-reinforcing

More users → more data → better algorithm → better service → more users → greater market power.

5. Main Forms of Machine-Directed Market Foreclosure

A. Algorithmic Self-Preferencing

A dominant platform may design an algorithm that systematically gives preferential treatment to its own products or services.

Example:

An online marketplace uses an AI ranking system that places its own products above competing products even when competing products perform better under objectively relevant criteria.

Competition concerns may include:

reduced visibility of competitors;

diversion of demand;

exclusion from important distribution channels;

reinforcement of platform dominance.

6. Algorithmic Exclusive Dealing

An AI platform may automatically encourage or enforce exclusive relationships.

For example:

An automated procurement system gives suppliers substantially better rankings if they sell exclusively through the platform.

The concern becomes stronger where:

the platform is dominant;

a large percentage of suppliers participate;

competitors cannot obtain equivalent distribution;

switching costs are high;

network effects reinforce the exclusion.

7. Algorithmic Tying and Bundling

A machine-controlled ecosystem may automatically combine products.

For example:

Access to a dominant AI model is automatically conditioned on purchasing the provider's cloud-computing service.

Possible competition concerns include:

leveraging dominance from one market into another;

raising competitors' costs;

reducing customer choice;

preventing competing suppliers from reaching customers.

8. Algorithmic Refusal of Access

A platform or infrastructure provider may use automated systems to determine who receives access.

Examples include:

refusing API access;

restricting cloud access;

blocking interoperability;

denying marketplace participation;

restricting data access;

automatically suspending competitors;

preventing integration with competing services.

This may become particularly important where the infrastructure is difficult to replicate.

9. Algorithmic Discrimination

Machine systems can apply different:

prices;

rankings;

commissions;

access conditions;

advertising opportunities;

visibility;

search results;

to competing businesses.

The key question is whether the differentiation has an objective and legitimate justification or instead produces exclusionary effects.

10. Data-Based Foreclosure

Data can function as a competitive input.

A dominant platform may collect:

customer data;

supplier data;

transaction data;

search data;

behavioural data;

pricing data;

performance data.

The platform may then use that information to improve its own competing service.

This can create a feedback loop:

Users → Data → Better Algorithm → Better Product → More Users → More Data

This can make market entry progressively more difficult.

11. AI-Powered Ranking Foreclosure

Search and ranking algorithms can determine commercial visibility.

Suppose an AI marketplace has 10,000 sellers but only the first 20 results receive substantial consumer attention.

If the platform manipulates ranking to disadvantage competitors, algorithmic ranking can become a form of market foreclosure.

Relevant factors include:

transparency;

ranking criteria;

market power;

discriminatory treatment;

duration;

coverage;

impact on competitors;

availability of alternative distribution channels.

12. Machine-Controlled Loyalty and Switching Costs

AI systems can make users increasingly dependent on one ecosystem.

Examples:

personalized recommendations;

proprietary data formats;

automated workflows;

AI assistants trained around one ecosystem;

incompatible APIs;

proprietary AI agents;

automatic subscription renewal;

personalized discounts unavailable elsewhere.

These mechanisms may make switching technically or economically difficult.

13. Network Effects and Machine Foreclosure

Machine-directed foreclosure becomes especially significant in markets with network effects.

For example:

More users → more data → better AI → better service → more users.

A competitor may therefore face a problem beyond ordinary entry costs.

It may need to overcome:

data advantages;

algorithmic advantages;

user networks;

developer networks;

distribution networks;

ecosystem integration.

This can produce self-reinforcing foreclosure.

14. At Least 6 Important Case Laws

The following cases are particularly useful. Some are direct authorities on exclusionary conduct, while others are analogous authorities whose principles can be applied to machine-directed foreclosure.

Case 1: United States v. Microsoft Corp., 253 F.3d 34 (D.C. Cir. 2001)

Facts

Microsoft was found to possess monopoly power in the market for Intel-compatible PC operating systems.

The case concerned Microsoft's conduct toward competing browsers, particularly Netscape.

Principle

The court examined whether Microsoft's contractual and technological restrictions unlawfully maintained monopoly power.

Relevance to Machine-Directed Foreclosure

This is one of the most important analogies for automated ecosystems.

Modern AI platforms can similarly control:

APIs;

operating environments;

application access;

interoperability;

distribution;

default settings.

The technological mechanism may change, but the competition question remains:

Is control over a critical technological ecosystem being used to exclude competing products?

Lesson

Technological integration does not automatically immunize exclusionary conduct from antitrust scrutiny.

15. Case 2: Google Shopping — European Commission / General Court Proceedings

Facts

The European Commission examined Google's treatment of competing comparison-shopping services in its search results.

The Commission found that Google systematically positioned and displayed its own comparison-shopping service more favourably than competing services.

The EU proceedings became a major authority concerning digital-platform self-preferencing.

Principle

A dominant digital platform's control over an important search/distribution channel can raise Article 102 TFEU concerns when that control disadvantages competing services.

Relevance

This is highly relevant to machine-directed foreclosure because ranking and visibility can be controlled through algorithms.

A modern AI marketplace could theoretically:

automatically rank its own AI products above competing products.

The relevant issue would be whether the algorithmic design produces exclusionary effects and whether there are objective justifications.

Lesson

Algorithmic control over visibility can become economically equivalent to control over distribution.

16. Case 3: Google Android, Case AT.40099

Facts

The European Commission examined Google's conduct concerning Android and several contractual arrangements involving mobile devices and applications.

The Commission considered whether Google's practices strengthened its position in general search and related markets.

Principle

Dominance in one technological ecosystem can potentially be leveraged into neighbouring markets through contractual and technological arrangements.

Relevance to Machine-Directed Foreclosure

AI ecosystems can similarly combine:

operating systems;

cloud;

AI models;

applications;

app stores;

advertising;

data;

devices.

An AI provider could potentially use control over one layer to disadvantage competitors at another layer.

Lesson

Vertical technological integration can create foreclosure risks where ecosystem control is leveraged across markets.

17. Case 4: Bronner v. Mediaprint, Case C-7/97

Facts

Bronner sought access to Mediaprint's newspaper home-delivery distribution system.

The issue concerned whether refusal of access to infrastructure controlled by another undertaking could constitute an abuse of dominance.

Principle

The Court of Justice established a demanding framework for compulsory access to infrastructure.

Among other considerations, the facility must be indispensable and duplication must not be realistically possible under the relevant circumstances.

Relevance to Machine-Directed Foreclosure

Modern machine-driven infrastructure may include:

cloud computing;

AI compute;

specialized datasets;

APIs;

digital identity infrastructure;

payment infrastructure;

platform interfaces.

Not every refusal to provide access constitutes unlawful foreclosure.

Lesson

Control of important technology does not automatically create an obligation to share it; indispensability and competitive effects matter.

18. Case 5: Commercial Solvents Corp. v. Commission, Joined Cases 6/73 and 7/73

Facts

Commercial Solvents was dominant in the production of certain raw materials and also participated in downstream markets.

It restricted supplies to a downstream competitor.

Principle

A dominant undertaking may abuse its position where it uses control over an upstream input to eliminate competition in a downstream market.

Relevance to Machine-Directed Foreclosure

Consider an AI company controlling a critical upstream input such as:

specialized AI compute;

an essential dataset;

an AI model;

a technical API.

If it competes downstream while controlling access to the upstream input, the possibility of foreclosure becomes important.

Example

Suppose Company A controls a critical AI API and also operates an AI application marketplace.

It could theoretically disadvantage competing applications by:

restricting API access;

increasing competitors' costs;

degrading technical performance;

giving its own applications superior access.

Lesson

Vertical control over a critical technological input can create downstream foreclosure risks.

19. Case 6: MOTOE v. Elliniko Dimosio, Case C-49/07

Facts

A Greek public body had regulatory responsibilities concerning motorcycling activities while also being involved in organizing such activities.

Principle

The Court considered the competition problems arising when an entity simultaneously possesses regulatory authority and participates in economic activity.

Relevance

Machine-controlled ecosystems can create a similar structural problem.

For example, a platform may:

establish the technical rules;

control access;

rank participants;

monitor compliance;

compete against those participants.

The concern is particularly strong where the platform effectively acts as both:

referee + competitor.

Lesson

Combining rule-making power with commercial participation can create structural foreclosure risks.

20. Case 7: United Brands Co. v. Commission, Case 27/76

Facts

United Brands was found to hold a dominant position in the relevant banana market.

The case became a foundational authority concerning dominance and abusive conduct.

Principle

Dominance itself is not prohibited; abusive exploitation or exclusionary conduct by a dominant undertaking is the competition-law concern.

Relevance

The principle is directly applicable to AI ecosystems.

A company may legitimately become technologically successful.

The competition concern arises where substantial market power is used to:

exclude competitors;

impose unfair conditions;

restrict access;

discriminate;

prevent effective competition.

Lesson

Market success is not itself unlawful; abusive exploitation of dominance is the critical issue.

21. Case 8: AKZO Chemie BV v. Commission, Case C-62/86

Facts

AKZO was a dominant undertaking and was accused of using pricing practices that disadvantaged a smaller competitor.

Principle

The case became an important authority on predatory pricing and the use of pricing strategies by dominant firms.

Relevance to Machine-Directed Foreclosure

AI systems can automatically set prices.

A dominant platform could theoretically use machine-controlled pricing to:

temporarily underprice competitors;

target competitors individually;

use data to identify vulnerable entrants;

coordinate discounts across an ecosystem.

Automated pricing does not remove competition-law responsibility.

Lesson

Automation does not transform potentially exclusionary pricing into lawful conduct.

22. Case 9: Ohio v. American Express Co., 585 U.S. 529 (2018)

Facts

American Express operated a two-sided transaction platform involving merchants and cardholders.

The Supreme Court examined the relevant market and the effects of AmEx's contractual restrictions.

Principle

In two-sided transaction platforms, competition analysis may need to consider both sides of the platform.

Relevance

This is extremely important for machine-driven platforms.

An AI marketplace may simultaneously serve:

consumers;

suppliers;

advertisers;

developers;

payment providers.

Foreclosure on one side may affect competition on the other.

Lesson

Machine-directed platform foreclosure must be analysed with the platform's interconnected sides in mind.

23. Case 10: Intel Corp. v. European Commission, Case C-413/14 P

Facts

The case concerned rebates offered by Intel to major computer manufacturers and a retailer.

Principle

The Court emphasized the importance of examining the actual or potential capability of the conduct to foreclose equally efficient competitors where such analysis is relevant.

Relevance

Modern machine-controlled rebate systems can dynamically determine:

discounts;

commissions;

advertising rates;

preferred placement;

supplier incentives.

Lesson

Automated commercial incentives should be assessed according to their competitive effects rather than merely their technical form.

24. Case 11: Eturas UAB v. Lietuvos Respublikos konkurencijos taryba, Case C-74/14

Facts

Eturas operated a common online travel-booking system. A system message communicated a restriction concerning discounts available through the platform.

Principle

The case addressed the evidentiary implications of communications occurring through an electronic platform and participation in an automated system.

Relevance

This case is particularly valuable for machine-directed markets because it demonstrates that competition-law analysis can extend to conduct implemented through software systems.

Modern systems may automatically transmit:

pricing rules;

discount restrictions;

market information;

supplier conditions.

Lesson

Competition law applies to electronically implemented coordination; software can be the mechanism through which competitive restrictions operate.

25. Algorithmic Foreclosure and Market Definition

Before determining foreclosure, competition authorities generally need to identify the relevant market or competitive constraint.

Possible markets include:

Product market

AI foundation models;

cloud computing;

AI applications;

online marketplaces;

digital advertising;

payment services;

logistics.

Geographic market

Could be:

national;

regional;

EU-wide;

global;

platform-specific.

Dynamic market definition

AI markets require particular attention to rapidly changing competitive conditions.

Today's separate market may become tomorrow's integrated ecosystem.

26. Market Power Indicators

Authorities may examine:

market share;

entry barriers;

network effects;

data advantages;

switching costs;

interoperability;

multi-homing;

technological superiority;

access to compute;

intellectual property;

user base;

ecosystem integration;

financial resources;

control over distribution.

A large market share is relevant but not automatically conclusive.

27. Foreclosure of Actual Competitors

Machine-directed conduct may harm existing competitors by:

reducing their rankings;

increasing their costs;

limiting access;

denying interoperability;

restricting data;

reducing customer visibility;

excluding them from distribution.

28. Foreclosure of Potential Competitors

Potential competition is particularly important in AI markets.

A large incumbent might acquire or suppress:

emerging AI startups;

competing algorithms;

alternative AI models;

new distribution technologies;

innovative infrastructure.

The concern is that the competitive harm may occur before the entrant becomes a major competitor.

29. Killer Acquisitions and Machine Ecosystems

AI markets may produce acquisitions involving companies with:

small current revenues;

valuable data;

innovative algorithms;

specialized engineers;

promising technology.

Competition authorities may therefore need to consider:

Who could this company become if it remained independent?

This is a potential-competition issue rather than simply a current-market-share issue.

30. Algorithmic Self-Learning and Competition Law

A difficult question arises where an AI system changes its behaviour without a specific human instruction.

For example:

Two autonomous pricing systems independently learn that maintaining higher prices increases profitability.

Possible questions include:

Was there an agreement?

Was there human communication?

Was the algorithm designed to facilitate coordination?

Did the undertaking knowingly deploy the system?

Can the conduct be attributed to the undertaking?

Is the outcome unilateral or coordinated?

Competition law must distinguish:

parallel algorithmic behaviour

from

unlawful coordination or exclusionary conduct.

31. Machine Foreclosure vs Efficient Automation

Automation is not inherently anti-competitive.

A company may legitimately use AI to:

reduce costs;

improve logistics;

personalize products;

improve fraud detection;

optimize inventory;

improve customer service.

The competition concern arises when automation is used to exclude competitors rather than compete on the merits.

32. Possible Legitimate Justifications

A firm may argue that automated restrictions are necessary for:

Security

Preventing fraud or cyberattacks.

Quality control

Maintaining technical or safety standards.

Privacy

Protecting confidential information or personal data.

Reliability

Ensuring stable system performance.

Consumer protection

Preventing misleading or harmful transactions.

Intellectual-property protection

Protecting legitimate proprietary technology.

Efficiency

Reducing costs or improving service quality.

The key question is usually whether the restriction is:

necessary + objectively justified + proportionate.

33. Competition-Law Risks

Machine-directed foreclosure can produce:

1. Entry barriers

New firms cannot obtain access to customers or infrastructure.

2. Reduced innovation

Competitors have fewer incentives or opportunities to innovate.

3. Higher prices

Reduced competition can ultimately increase prices.

4. Reduced quality

Competitive pressure may weaken.

5. Reduced consumer choice

Consumers become dependent on one ecosystem.

6. Market tipping

A platform may become extremely difficult to challenge.

7. Technological dependency

Competitors may become dependent upon infrastructure controlled by a dominant firm.

34. India Perspective

In India, the principal framework is the Competition Act, 2002.

Machine-directed foreclosure may potentially implicate:

Section 3

Prohibits agreements having or likely to have appreciable adverse effect on competition.

Relevant conduct may include:

market allocation;

exclusive arrangements;

discriminatory agreements;

certain vertical restraints.

Section 4

Concerns abuse of dominant position.

Possible theories include:

unfair or discriminatory conditions;

unfair pricing;

limiting markets;

denial of market access;

leveraging dominance;

tying/bundling.

Digital Competition

For algorithmic platforms, the Competition Commission of India may need to consider:

network effects;

data;

platform dependency;

self-preferencing;

exclusive arrangements;

interoperability;

switching costs;

algorithmic discrimination.

35. UAE Perspective

In the UAE, competition issues involving machine-directed foreclosure can be analysed principally through the federal competition framework, including the Federal Law No. 4 of 2012 on Regulation of Competition, together with applicable regulations and sector-specific rules.

Relevant concepts include:

restrictive agreements;

abuse of dominant position;

market power;

economic concentration;

exclusionary conduct;

technological barriers;

access to important infrastructure.

For technology-intensive markets, the factual analysis may need to consider whether automated conduct produces a genuine restriction of competition rather than assuming that technological sophistication itself is unlawful.

36. EU Perspective

EU competition law is particularly relevant because Article 101 TFEU addresses anti-competitive agreements, while Article 102 addresses abuse of dominance.

Digital-platform cases demonstrate increasing attention to:

self-preferencing;

platform leverage;

ecosystem foreclosure;

tying;

exclusionary contracts;

digital distribution.

The Digital Markets Act also provides an important ex-ante framework for designated gatekeepers, operating alongside traditional competition law.

37. United States Perspective

In the United States, machine-directed foreclosure may potentially involve:

Sherman Act §1;

Sherman Act §2;

Clayton Act merger provisions;

FTC Act;

state antitrust statutes.

The analytical distinction between:

agreement-based restrictions

and

unilateral exclusionary conduct

remains important.

38. Remedies

Where unlawful machine-directed foreclosure is established, remedies may include:

Behavioral remedies

stopping discriminatory algorithms;

changing ranking systems;

removing exclusivity;

ending restrictive contracts;

permitting interoperability.

Structural remedies

In exceptional circumstances:

divestiture;

separation of business units;

structural separation of infrastructure and downstream operations.

Data remedies

data portability;

access mechanisms;

interoperability;

restrictions on combining certain datasets.

Algorithmic remedies

independent audits;

monitoring;

explainability requirements;

testing;

compliance logs;

human oversight.

Merger remedies

divestiture;

licensing;

interoperability commitments;

access obligations.

39. Compliance Framework for Businesses

Companies using AI should ask:

Is the company dominant?

What relevant market is affected?

Does the algorithm control access to customers?

Does it disadvantage competitors?

Does it favour the company's own products?

Are competitors dependent on the infrastructure?

Are there exclusivity provisions?

Can customers multi-home?

Can customers switch easily?

Is interoperability available?

Is the restriction objectively justified?

Is it proportionate?

Does the algorithm use competitor-sensitive data?

Could the system facilitate coordination?

Are algorithmic decisions auditable?

Is there human competition-law oversight?

40. Practical Example

Imagine AutoMarket AI, a dominant autonomous marketplace.

It:

ranks products;

sets advertising prices;

recommends suppliers;

processes payments;

controls logistics;

provides AI purchasing agents;

collects consumer data.

AutoMarket AI begins ranking its own products first.

It also:

reduces the visibility of competing suppliers;

gives its own products preferential advertising;

limits API access for competitors;

requires suppliers to use its payment system;

makes switching technically difficult.

The competition-law analysis would examine:

Market power → platform control → algorithmic conduct → competitor disadvantage → foreclosure → consumer effects → efficiency justification → proportionality → remedy.

The fact that an AI system performed the conduct does not by itself determine legality.

41. Key Distinction

42. Overall Legal Test

A useful analytical sequence is:

Step 1 — Identify the market

What product, service, technology or infrastructure is involved?

Step 2 — Identify market power

Does the undertaking possess substantial power?

Step 3 — Identify the machine-controlled conduct

What exactly does the algorithm or automated system do?

Step 4 — Identify the competitive restriction

Does it exclude, disadvantage or marginalize competitors?

Step 5 — Assess foreclosure

How much of the market is affected?

Step 6 — Examine causation

Did the machine-directed conduct materially contribute to the exclusion?

Step 7 — Examine efficiencies

Are there legitimate technical, security, quality or efficiency explanations?

Step 8 — Apply proportionality

Could the legitimate objective be achieved through a less restrictive mechanism?

Step 9 — Consider long-term effects

Does the conduct reinforce network effects, entry barriers or technological dependency?

Step 10 — Select remedies

Use behavioral, interoperability, access, monitoring or structural remedies as appropriate.

43. Important Legal Principle

The central principle is:

Competition law regulates economic effects and competitive conduct, not merely the identity of the decision-maker.

Therefore, replacing a human decision-maker with an algorithm does not automatically remove competition-law responsibility.

44. Conclusion

Machine-directed market foreclosure represents the intersection of competition law, artificial intelligence, digital platforms and technological infrastructure.

The major concern is not automation itself. The concern is the use of automated systems by firms possessing substantial market power to:

exclude competitors;

restrict market access;

favour affiliated products;

control essential inputs;

increase switching costs;

exploit data advantages;

reinforce network effects;

prevent future competition.

The traditional cases involving Microsoft, United Brands, AKZO, Bronner, Commercial Solvents, MOTOE, Google Shopping, Google Android, American Express, Intel and Eturas provide useful legal principles even though most predate modern generative AI.

The future challenge is therefore to apply established competition principles to increasingly automated markets without treating every algorithmic decision as unlawful.

Quick Revision Formula

Machine-Directed Market Foreclosure =

**Market Power

Algorithmic Control

Data/Infrastructure Advantage

Exclusionary Conduct

Entry Barriers

Network Effects

Competitor Foreclosure

Reduced Contestability

Efficiency Assessment

Proportionate Remedies**

One-Line Exam Definition

Machine-directed market foreclosure is the use or control of AI, algorithms, automated systems or machine-operated infrastructure by an undertaking to restrict competitors' effective access to customers, inputs, distribution channels, data or technology, thereby potentially weakening competition or reinforcing market power.