Competition Law and Competition Governance in Machine-Mediated Markets

1. Introduction

Machine-mediated markets are markets in which important competitive decisions are made, assisted, or substantially influenced by algorithms, artificial intelligence, machine learning, automated pricing systems, recommendation engines, ranking systems, digital platforms, autonomous agents, or other computational mechanisms.

In traditional markets, firms and human decision-makers directly determine prices, output, quality, access and distribution. In machine-mediated markets, these functions may be delegated to software. This creates a fundamental competition-law question:

When a machine makes or influences a competitive decision, the underlying conduct remains subject to competition law.

The technological form of the conduct does not by itself determine its legality. For example, an algorithm can facilitate price coordination, exclusionary conduct, self-preferencing, discriminatory access, information exchange or market allocation.

Recent enforcement demonstrates this increasingly clearly. In the United States, the DOJ has pursued algorithmic-pricing cases involving RealPage, while the FTC and DOJ have stated that competitors cannot evade antitrust law merely because coordination is implemented through a common algorithm.

2. Meaning of Machine-Mediated Markets

A machine-mediated market exists where computational systems materially influence the interaction between buyers and sellers.

Examples include:

1. Algorithmic pricing
   • Dynamic pricing
   • Revenue-management software
   • Surge pricing
   • Automated price matching
2. Algorithmic ranking
   • Search results
   • Marketplace rankings
   • Product recommendations
   • App-store rankings
3. Automated allocation
   • Advertising auctions
   • Digital advertising inventory
   • Delivery allocation
   • Ride-hailing assignments
4. AI-enabled decision-making
   • Credit decisions
   • Insurance pricing
   • Fraud detection
   • Personalized offers
5. Platform-mediated markets
   • E-commerce marketplaces
   • App stores
   • Digital advertising exchanges
   • Online travel platforms
6. Autonomous commercial agents
   • AI purchasing agents
   • Automated negotiation systems
   • Autonomous trading systems
   • Machine-to-machine transactions

The competition-law problem becomes particularly significant when multiple competitors rely upon the same machine, data pool, platform or optimization mechanism.

3. Competition Governance

Competition governance refers to the institutional and technological mechanisms used to ensure that machine-mediated markets remain competitive.

It therefore extends beyond conventional antitrust enforcement.

Major components

The central objective is not to prohibit the use of machines. Rather, it is to ensure that automation does not eliminate competitive independence.

4. Major Competition-Law Issues

A. Algorithmic Price Fixing

This is one of the most important problems.

Suppose competing firms independently provide their pricing information to the same algorithm. The algorithm analyses the information and recommends prices that cause competitors to converge.

The legal question is not simply:

"Did a human competitor call another competitor?"

Instead, regulators may examine whether there was an arrangement, understanding, information exchange or other mechanism that replaced independent competitive decision-making.

The DOJ has expressly argued in the RealPage litigation that algorithmic implementation does not immunize price-fixing conduct from antitrust law.

Relevant factors

• Common algorithm
• Competitor participation
• Exchange of non-public information
• Knowledge that rivals are using the same system
• Pricing recommendations
• Rules designed to reduce price competition
• Restrictions on discounts
• Monitoring of competitor behaviour

5. Hub-and-Spoke Algorithmic Coordination

Machine-mediated markets can create a modern hub-and-spoke structure.

Traditional structure

Competitor A ↔ Competitor B ↔ Competitor C

Algorithmic structure

Competitor A → Algorithm Provider ← Competitor B
Competitor C → Algorithm Provider

The algorithm provider becomes the hub, while competing businesses become the spokes.

The risk increases where:

• the hub receives competitively sensitive information from competitors;
• the same algorithm is used by competing firms;
• competitors know that rivals participate;
• the algorithm generates coordinated recommendations;
• the system discourages independent pricing.

The FTC and DOJ's 2024 statement of interest in Cornish-Adebiyi v. Caesars Entertainment emphasized that direct communication between competitors is not necessarily required to plead an unlawful agreement and that common algorithmic pricing mechanisms may raise traditional price-fixing concerns.

6. Algorithmic Tacit Coordination

A more difficult problem occurs when algorithms independently learn that coordination is profitable.

For example:

• Firm A's algorithm raises price.
• Firm B's algorithm detects the change.
• B's algorithm raises price.
• A's algorithm observes B's response.
• Both algorithms learn that aggressive price competition is undesirable.

There may be no explicit agreement.

This creates a distinction between:

Explicit algorithmic collusion

Human actors intentionally design or use an algorithm to coordinate.

Tacit algorithmic coordination

Algorithms independently learn strategies that produce parallel outcomes.

Competition law traditionally distinguishes unlawful agreement from lawful conscious parallelism. Machine learning makes that distinction more difficult because autonomous systems can discover strategies that their designers did not expressly program.

7. Information Exchange

Machine-mediated markets can make information exchange much faster and more comprehensive.

Algorithms can process:

• prices;
• inventory;
• demand;
• customer behaviour;
• discounts;
• future pricing;
• capacity;
• geographic information;
• competitor activity.

A system can therefore transform apparently ordinary data into a mechanism for coordinating competitors.

The RealPage litigation illustrates the concern: the DOJ alleged that competing landlords supplied competitively sensitive information to a common pricing system that generated pricing recommendations.

8. Algorithmic Self-Preferencing

A dominant digital platform may control the algorithm that determines which businesses receive visibility.

For example:

Platform → controls ranking algorithm → ranks own service above rivals

Potential competition concerns include:

• preferential ranking;
• preferential search placement;
• discriminatory recommendation;
• preferential advertising allocation;
• manipulation of marketplace visibility.

The European Commission has specifically addressed Google's self-preferencing under the Digital Markets Act, including preferential treatment of Google's own services in search results.

9. Algorithmic Discrimination

Machine-mediated markets can discriminate between:

• competitors;
• consumers;
• suppliers;
• geographic areas;
• products;
• distribution channels.

Competition law may become relevant where discrimination is used by a dominant undertaking to exclude rivals.

For example, a dominant marketplace might algorithmically:

1. reduce a rival's visibility;
2. increase its own product ranking;
3. restrict access to valuable consumer data;
4. manipulate recommendation systems;
5. impose discriminatory terms.

The legal analysis would ordinarily consider dominance, foreclosure, effects on competition and objective justification, rather than treating every algorithmic difference as unlawful.

10. Algorithmic Exclusion

A dominant undertaking can potentially use an algorithm to implement exclusionary strategies.

Possible mechanisms include:

• automatic demotion;
• discriminatory access;
• algorithmic delisting;
• refusal to display competitors;
• interoperability restrictions;
• tying;
• loyalty mechanisms;
• exclusionary ranking;
• discriminatory API access.

This makes the algorithm both a commercial tool and a potential instrument of market power.

11. Data as a Competitive Asset

Machine-mediated markets depend heavily on data.

A firm possessing a unique dataset may obtain advantages in:

• prediction;
• personalization;
• pricing;
• recommendation;
• fraud detection;
• machine learning;
• advertising.

Competition authorities may therefore examine:

Who controls the data necessary to compete?

Potential concerns include:

• data foreclosure;
• refusal of access;
• discriminatory data access;
• exclusive data agreements;
• data accumulation through mergers;
• tying data access to another service;
• restrictions on portability.

12. Network Effects

Machine-mediated markets frequently exhibit strong network effects.

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

This can produce a feedback loop:

Users → Data → Machine Learning → Better Service → More Users

The same feedback loop can become a barrier to entry.

A new entrant may possess an excellent algorithm but lack:

• historical data;
• users;
• distribution;
• infrastructure;
• interoperability;
• training data.

Consequently, competition authorities may need to distinguish between legitimate innovation advantages and strategic exclusion.

13. Switching Costs and Algorithmic Lock-In

Consumers and businesses can become dependent on a machine-mediated ecosystem.

Examples include:

• proprietary APIs;
• proprietary data formats;
• machine-learning models;
• platform-specific reputation scores;
• algorithmic recommendations;
• cloud infrastructure;
• automated workflow systems.

The longer a firm remains in the ecosystem, the greater the cost of switching.

This can reinforce market power even where the initial service was offered at low or zero monetary price.

14. Merger Control

Machine-mediated markets create new merger-control questions.

A transaction may involve:

• acquisition of an AI company;
• acquisition of a data provider;
• acquisition of an algorithmic marketplace;
• acquisition of an advertising exchange;
• acquisition of an autonomous-agent platform.

Traditional turnover-based thresholds may not always capture strategically important acquisitions.

Authorities may therefore examine:

• data assets;
• innovation competition;
• potential competition;
• access to algorithms;
• interoperability;
• vertical foreclosure;
• ecosystem effects;
• control of complementary technologies.

15. Six Important Case Laws / Enforcement Matters

1. United States v. RealPage Inc. — United States

The DOJ sued RealPage over alleged algorithmic coordination in rental housing.

The government alleged that landlords supplied competitively sensitive information to RealPage and used its pricing system to generate rental-price recommendations.

The case is significant because it directly addresses the proposition that algorithmic implementation does not remove traditional antitrust liability. The DOJ alleged violations of Sections 1 and 2 of the Sherman Act.

Principle

Competitors cannot necessarily avoid price-fixing rules by outsourcing pricing decisions to a common algorithm.

2. Cornish-Adebiyi v. Caesars Entertainment, Inc. — United States

This litigation concerned allegations that hotels used a common algorithmic pricing system.

The FTC and DOJ filed a statement of interest explaining that algorithmic coordination can potentially constitute unlawful price fixing and that direct competitor-to-competitor communication is not necessarily indispensable to establishing an agreement.

The agencies also emphasized that common algorithms can facilitate coordination where competitors provide information to a shared algorithmic mechanism.

Principle

A common algorithm can become a mechanism through which competing firms coordinate pricing.

3. United States v. Topkins — United States

The Topkins matter involved online sellers who allegedly used algorithms to implement an agreement concerning prices for posters sold online.

The matter is historically important because it demonstrated that conventional cartel principles can apply to computer-programmed pricing.

Principle

Programming a price-fixing agreement into software does not transform the underlying cartel into lawful conduct.

It is an early illustration of the transition from human-to-human coordination toward machine-assisted coordination.

4. Google Search (Shopping) — European Union

In Google Search (Shopping), the European Commission found that Google abused a dominant position by giving preferential treatment to its own comparison-shopping service in search results.

The EU courts subsequently considered the legal framework governing this conduct, including the relationship between Google's dominance, its search-ranking mechanisms and foreclosure effects.

Principle

Algorithmic ranking can constitute an instrument of exclusion where a dominant platform uses it to favour its own service and disadvantage competing services.

This is especially important for machine-mediated markets because ranking algorithms can determine access to consumers.

5. Google Search (Android) — European Union

The Google Android proceedings concerned Google's contractual and platform practices surrounding Android, including restrictions connected with the distribution of Google Search and related services.

The case demonstrates how control over an operating-system ecosystem can interact with:

• default settings;
• distribution;
• search;
• app ecosystems;
• network effects.

Principle

Control over a technological ecosystem can be used in ways that reinforce dominance in adjacent markets.

6. United States v. Google LLC — Search Distribution

The U.S. Department of Justice's search case against Google examines Google's conduct concerning distribution arrangements and the maintenance of its position in general search.

The litigation illustrates a broader competition-governance problem: when a dominant technological intermediary controls the mechanisms through which users reach competing services, contractual and technological arrangements can reinforce market power.

Principle

Distribution agreements, default mechanisms and technological ecosystems can be important instruments of maintaining market power.

16. Additional Important Authorities

Several earlier competition cases remain useful when analysing machine-mediated markets because they establish principles that can be applied to algorithmic environments.

United States v. Apple Inc.

Relevant to:

• platform restrictions;
• ecosystem control;
• interoperability;
• exclusion of rivals.

FTC v. Amazon.com, Inc.

Relevant to:

• marketplace power;
• seller restrictions;
• pricing practices;
• platform governance.

Ohio v. American Express

Relevant to:

• two-sided platforms;
• platform economics;
• network effects;
• market definition.

FTC v. Facebook

Relevant to:

• data-driven platforms;
• network effects;
• exclusion;
• acquisition of potential competitors.

These authorities show that machine-mediated markets do not require an entirely new competition-law system. Instead, existing doctrines increasingly have to be applied to data, algorithms and technological ecosystems.

17. Competition Governance Framework

A comprehensive governance framework can be divided into six levels.

Level 1 — Ex ante design

Competition considerations should be incorporated when algorithms are designed.

Questions include:

• Does the algorithm use competitor-sensitive information?
• Can it facilitate coordination?
• Does it discriminate against rivals?
• Does it favour the platform's own products?
• Does it create unnecessary switching costs?

Level 2 — Data governance

Businesses should determine:

• what data are collected;
• who owns the data;
• who receives access;
• whether competitor data are pooled;
• whether information is competitively sensitive.

Level 3 — Algorithmic auditing

Audits can examine:

• pricing outputs;
• ranking;
• recommendations;
• discriminatory effects;
• exclusionary rules;
• changes in model behaviour.

Level 4 — Human oversight

Important competitive decisions should have accountable human oversight.

This is particularly important where machine-learning systems modify their behaviour over time.

Level 5 — Regulatory monitoring

Competition authorities may monitor:

• algorithmic markets;
• digital platforms;
• AI markets;
• pricing systems;
• mergers involving data;
• platform ecosystems.

Level 6 — Remedial intervention

Possible remedies include:

• prohibition of information sharing;
• algorithmic restrictions;
• interoperability;
• data access;
• non-discrimination;
• structural remedies;
• compliance monitoring.

The DOJ's RealPage-related settlements illustrate the emergence of highly specific algorithmic remedies, including restrictions on use of competitively sensitive data and monitoring requirements.

18. Evidence and Algorithmic Investigations

Machine-mediated competition cases create unusual evidentiary problems.

Authorities may need to examine:

Source code

To determine how the algorithm operates.

Training data

To determine what information shaped the model.

Input data

To determine whether competitors supplied sensitive information.

Model outputs

To establish the actual market effect.

Logs

To reconstruct algorithmic decisions.

Version histories

To determine when potentially problematic features were introduced.

Internal communications

To determine whether humans intentionally designed the system to coordinate or exclude rivals.

A/B testing

To determine whether algorithmic changes affected competitors or consumers.

Thus, competition investigations increasingly require technical forensic capabilities in addition to conventional documentary evidence.

19. Explainability and Accountability

A major governance problem is the black-box algorithm.

Suppose an AI system produces the following recommendation:

"Increase price by 18%."

The regulator must ask:

• Why 18%?
• Which data were used?
• Were competitor prices used?
• Were competitor-specific data used?
• Was the model trained on confidential information?
• Did the system intentionally reduce price competition?
• Did the company know how the model arrived at the recommendation?

Consequently, explainability becomes a competition-governance issue.

20. Autonomous AI Agents and Future Competition Law

The next stage is likely to involve AI agents that negotiate directly with other AI agents.

For example:

Consumer AI Agent → searches market → compares products → negotiates price

while

Seller AI Agent → evaluates demand → changes price → negotiates automatically

This could produce markets where humans do not directly negotiate with each other.

Competition law may therefore have to address:

1. autonomous price adjustments;
2. machine-to-machine negotiation;
3. algorithmic market allocation;
4. autonomous purchasing;
5. automated switching;
6. machine-generated contracts;
7. autonomous collusion;
8. algorithmic exclusion.

The central question will remain whether competitive independence has been preserved.

21. Difference Between Legitimate Automation and Anticompetitive Automation

The use of AI or algorithms does not itself create an antitrust violation. The relevant question is the competitive conduct and its effects.

22. Indian Competition-Law Perspective

For India, the principal framework is the Competition Act, 2002, administered by the Competition Commission of India.

Machine-mediated markets can implicate:

Section 3

Anti-competitive agreements, including:

• price fixing;
• market allocation;
• bid rigging;
• information exchange;
• vertical restraints.

Section 4

Abuse of dominant position, including:

• discriminatory conditions;
• discriminatory prices;
• denial of market access;
• tying;
• leveraging;
• exclusionary conduct.

Sections 5 and 6

Combinations and merger control.

This framework is technologically neutral enough to address many algorithmic practices without requiring that the legislation expressly use the word "algorithm."

23. Key Doctrinal Challenges

1. Attribution

Who is responsible for an autonomous algorithm?

• developer?
• platform?
• user?
• algorithm provider?
• business using the software?

2. Intent

How should competition law treat conduct where the algorithm learns behaviour rather than receiving an explicit instruction?

3. Agreement

When does simultaneous algorithmic behaviour amount to an agreement?

4. Market definition

Should an AI platform be considered:

• a separate market;
• part of a broader digital market;
• a multi-sided platform;
• an ecosystem?

5. Dynamic markets

Traditional market-share analysis may become outdated rapidly because technological markets evolve quickly.

6. Innovation

Aggressive algorithmic optimization can simultaneously produce:

• efficiency;
• lower costs;
• innovation;

while potentially producing:

• foreclosure;
• concentration;
• exclusion.

Competition analysis therefore needs to distinguish legitimate innovation from anticompetitive conduct.

24. Emerging Governance Model

A useful model is:

DATA → ALGORITHM → DECISION → MARKET EFFECT → COMPETITION REVIEW

At every stage, regulators can ask:

Data

Was the information lawfully obtained?

Algorithm

Does the system incorporate anticompetitive rules?

Decision

Does it eliminate independent competitive decision-making?

Market effect

Does it foreclose competitors or facilitate coordination?

Governance

Are transparency, auditability and accountability adequate?

25. Core Principles

The emerging law of machine-mediated competition can be summarized through ten principles:

1. Technology neutrality — machines do not receive an antitrust exemption.
2. Competitive independence — competitors should retain independent decision-making.
3. Data sensitivity — competitively sensitive information requires careful treatment.
4. Algorithmic accountability — businesses remain responsible for systems they deploy.
5. Non-discrimination — dominant platforms should not unjustifiably discriminate against rivals.
6. Interoperability — technical architecture can affect competitive access.
7. Transparency — important algorithmic decisions should be capable of regulatory examination.
8. Auditability — logs, model histories and data provenance are increasingly important.
9. Innovation protection — competition law should not prohibit legitimate technological efficiency.
10. Human responsibility — delegation to software does not necessarily eliminate legal responsibility.

26. Conclusion

Machine-mediated markets represent a transformation in the mechanics of competition, rather than the disappearance of competition law.

Algorithms can improve markets by reducing transaction costs, improving matching, forecasting demand and enabling dynamic allocation. At the same time, they can facilitate coordination, amplify market power, discriminate against rivals, create technological barriers to entry and allow dominant platforms to control access to consumers.

The emerging regulatory approach therefore focuses on the conduct and competitive consequences behind the technology.

The RealPage proceedings are particularly significant because they demonstrate that authorities are prepared to apply traditional antitrust concepts to algorithmic pricing and shared computational systems. Similarly, the Google proceedings demonstrate how ranking, distribution and platform architecture can become central competition-law questions in machine-mediated ecosystems.

Thus, the central proposition for competition governance is:

A machine may make the competitive decision, but the competitive consequences remain subject to competition law.