Competition Law and Machine-Administered Competition Frameworks

1. Introduction

Machine-administered competition frameworks refer to competition systems in which algorithms, artificial intelligence (AI), automated platforms, autonomous agents, and computational tools perform an increasing part of the functions traditionally performed by human businesses, regulators, or market intermediaries.

The concept has two related dimensions:

1. Machines administering markets — algorithms determine prices, rankings, access, matching, allocation and transactions.
2. Machines assisting competition enforcement — regulators use AI and computational systems to detect cartels, analyze mergers, identify exclusionary conduct and monitor markets.

Thus, the central question is:

How can competition law preserve effective competition when both market conduct and competition administration increasingly depend upon machines?

There is not yet a universally recognized independent body of law called “machine-administered competition law.” It is better understood as a developing application of established competition principles to automated and AI-driven markets.

2. Meaning of Machine-Administered Competition Frameworks

A machine-administered competition framework can be defined as:

A legal, economic and technological framework in which automated systems substantially participate in the creation, operation, monitoring or enforcement of competitive conditions in markets.

It may involve:

• algorithmic pricing;
• automated procurement;
• AI-based market allocation;
• autonomous purchasing agents;
• automated platform rankings;
• machine-generated contracts;
• algorithmic merger screening;
• cartel detection;
• AI-assisted regulatory investigations;
• automated compliance systems; and
• computational market monitoring.

3. Traditional Competition Framework

Traditional competition law generally follows:

Human firms → human decisions → market conduct → legal investigation → human enforcement

Examples include:

• companies fixing prices;
• managers agreeing to divide markets;
• firms entering exclusive contracts;
• directors approving mergers;
• regulators investigating documents.

Machine-administered competition may produce:

Algorithm → automated decision → market outcome → computational evidence → regulatory analysis

This creates new legal questions concerning:

• responsibility;
• causation;
• intent;
• transparency;
• evidence;
• accountability; and
• judicial review.

4. Two Levels of Machine Administration

A. Market Administration

Machines can administer commercial activity by determining:

• prices;
• supply;
• demand;
• advertising;
• ranking;
• product recommendations;
• inventory;
• procurement;
• logistics;
• credit allocation.

B. Regulatory Administration

Competition authorities can use machines to:

• identify suspicious pricing patterns;
• monitor markets;
• screen mergers;
• analyze communications;
• detect cartel indicators;
• examine market concentration;
• predict competitive risks;
• process large datasets.

The second category is particularly important because the competition regulator itself may become technologically dependent on AI.

5. Objectives of Machine-Administered Competition Law

A sound framework should seek to preserve:

5.1 Competitive rivalry

Automation should not eliminate effective competition.

5.2 Market access

New entrants should have meaningful opportunities to compete.

5.3 Consumer choice

Consumers should not be forced into one technological ecosystem.

5.4 Innovation

Competition law should preserve incentives for technological development.

5.5 Accountability

Important automated decisions should remain legally attributable to identifiable economic actors.

5.6 Transparency

Authorities and affected parties should have sufficient information to challenge potentially unlawful automated conduct.

6. Algorithmic Pricing

One of the most important applications of machine administration is automated pricing.

A company may deploy an algorithm that:

1. observes market conditions;
2. observes competitors' prices;
3. forecasts demand;
4. changes prices;
5. evaluates the results;
6. continuously learns.

This can generate very rapid market responses.

The problem arises when algorithms facilitate coordination between competitors.

7. Machine-Assisted vs Machine-Generated Collusion

These situations should be distinguished.

1. Human agreement + machine implementation

Competitors agree to fix prices and use algorithms to implement the agreement.

This is the most straightforward case.

2. Common algorithm

Several competitors use the same pricing system.

The system may facilitate parallel pricing.

3. Algorithmic monitoring

Each company uses an independent algorithm to monitor competitors.

4. Autonomous coordination

Algorithms independently learn that coordinated pricing is profitable.

The last situation raises particularly difficult questions regarding:

• agreement;
• intention;
• attribution;
• foreseeability; and
• liability.

8. Machine Administration and Market Power

An automated system can itself become an instrument of market power.

A dominant platform may use algorithms to determine:

• which suppliers appear first;
• which products are recommended;
• which sellers receive access;
• which advertisements receive visibility;
• which applications are permitted.

Thus:

Algorithmic control can become a mechanism through which market power is exercised.

9. Automated Ranking and Self-Preferencing

Suppose a dominant platform operates:

• a marketplace;
• an AI assistant; and
• its own competing products.

Its algorithm may systematically prioritize its own products.

The competition issue is not simply that an algorithm made the decision.

The legal questions include:

• Who designed the algorithm?
• What objective was it given?
• Were affiliated products treated differently?
• Was the conduct capable of excluding competitors?
• Did the conduct affect competition?

10. Machine-Administered Access

A platform may automatically decide:

• who can enter;
• who can sell;
• which applications can operate;
• which APIs can be used;
• which users can access certain services.

Automated access rules can increase efficiency.

However, if a dominant undertaking uses them to exclude competitors, they can become relevant to competition law.

11. Automated Merger Screening

Competition authorities may increasingly use AI to screen proposed mergers.

An algorithm could analyze:

• market shares;
• transaction value;
• historical acquisitions;
• patent portfolios;
• customer overlap;
• product pipelines;
• technology;
• innovation indicators.

This could help identify potentially problematic transactions.

However, AI screening should not automatically determine the legal outcome.

Human and institutional review remains important because merger analysis involves:

• economic judgments;
• legal interpretation;
• evidence;
• uncertainty about future markets.

12. AI-Based Cartel Detection

Competition authorities can use algorithms to identify unusual patterns such as:

• simultaneous price movements;
• repeated bidding patterns;
• suspicious tender rotations;
• geographic allocation;
• unusual market stability;
• communication patterns.

For example:

Firm A → ₹100

Firm B → ₹101

Firm C → ₹99

If the pattern repeats in suspicious circumstances, automated systems can flag it for investigation.

But:

A statistical anomaly is evidence for investigation, not automatically proof of a cartel.

13. Machine Learning and Evidence

Machine-administered competition enforcement can process enormous quantities of:

• emails;
• contracts;
• invoices;
• transaction data;
• pricing information;
• internal communications;
• source-code records;
• algorithmic logs.

This can substantially increase investigative capacity.

However, regulators must ensure:

• evidentiary reliability;
• reproducibility;
• procedural fairness;
• confidentiality;
• protection of legally privileged material;
• explainability.

14. Due Process

Automated enforcement creates an important procedural question:

Can a company effectively challenge a competition decision if the regulator cannot explain how its AI reached the conclusion?

A machine-generated risk score should not automatically become a legal finding.

Affected businesses may need access to sufficient information to understand:

• the evidence relied upon;
• relevant methodology;
• assumptions;
• limitations;
• reasons for the decision.

15. Human Oversight

A strong framework should maintain human oversight.

A useful model is:

Machine detection → Human investigation → Legal analysis → Institutional decision → Judicial review

rather than:

Machine detection → Automatic punishment

The first approach preserves technological efficiency while maintaining legal accountability.

16. Case Law

There is no established body of reported cases specifically dealing with machine-administered competition frameworks. The following cases provide important competition-law principles that are relevant by analogy.

17. Case 1 — United States v. Microsoft Corp.

253 F.3d 34 (D.C. Cir. 2001)

Facts

Microsoft possessed substantial power in PC operating systems and engaged in conduct involving Internet Explorer and the Windows platform.

Principle

The case examined:

• platform power;
• network effects;
• exclusionary conduct;
• technological integration;
• barriers to entry.

Relevance

Machine-administered markets may similarly involve platforms controlling:

• AI systems;
• operating environments;
• application ecosystems;
• APIs;
• data.

An algorithmic platform can become a powerful intermediary between competitors and consumers.

Lesson

Control over a technological platform can become an important source of market power when it is used to restrict competitive opportunities.

18. Case 2 — United States v. Terminal Railroad Association

224 U.S. 383 (1912)

Principle

The case concerned strategically important transportation infrastructure and access to facilities necessary for effective competition.

Relevance to machine administration

Modern competition may depend on infrastructure such as:

• cloud computing;
• AI processors;
• data centers;
• communications networks.

If automated markets depend upon infrastructure controlled by one undertaking, access conditions may become competitively significant.

Lesson

Control over critical infrastructure can affect competition in downstream markets.

19. Case 3 — United Brands v. Commission

Case 27/76

Principle

The case established important principles concerning:

• dominance;
• market definition;
• independent commercial behavior;
• abuse of dominant position.

Machine relevance

A dominant machine platform may have substantial ability to act independently because rivals cannot easily reproduce its:

• data;
• technology;
• infrastructure;
• network;
• user base.

However, dominance alone does not establish an infringement.

Lesson

The central competition question is how market power is exercised.

20. Case 4 — Google Shopping

Google and Alphabet v. Commission, Case C-48/22 P

Competition issue

The case concerned Google's treatment of its comparison-shopping service within its general search ecosystem.

Relevance

AI systems increasingly perform functions similar to search engines.

An AI assistant may determine:

• which products appear;
• which supplier is recommended;
• which service is highlighted.

If the dominant platform systematically favors its own affiliated services, algorithmic ranking can become relevant to competition law.

Lesson

Automated ranking can influence competitive access to consumers.

21. Case 5 — Intel v. Commission

Intel Corp. v. Commission, Case C-413/14 P

Principle

The litigation concerned exclusionary rebates by a dominant undertaking and the assessment of their competitive effects.

Machine-administered relevance

Automated systems can determine:

• discounts;
• rebates;
• cloud credits;
• supplier incentives;
• preferential access.

A machine may administer these arrangements at enormous scale.

The fact that an algorithm calculates the discount does not remove the need for competition-law analysis.

Lesson

Automated commercial incentives must still be assessed according to their competitive effects.

22. Case 6 — Aspen Skiing Co. v. Aspen Highlands Skiing Corp.

472 U.S. 585 (1985)

Principle

The case concerned a dominant firm's withdrawal from cooperation with a rival in circumstances relevant to exclusionary conduct.

Machine relevance

A dominant platform may automatically terminate:

• API access;
• interoperability;
• technical integration;
• distribution arrangements.

Such conduct is not automatically unlawful.

However, where an established course of cooperation is strategically terminated to exclude a rival, competition concerns may arise depending on the circumstances.

Lesson

Automation does not eliminate the legal significance of strategically exclusionary conduct.

23. Case 7 — Verizon Communications Inc. v. Trinko

540 U.S. 398 (2004)

Principle

The U.S. Supreme Court emphasized that competition law generally does not impose an unrestricted duty upon firms to share their assets with competitors.

Machine relevance

AI companies may possess:

• proprietary models;
• proprietary datasets;
• algorithms;
• computing infrastructure;
• APIs.

A machine-administered competition framework must therefore avoid converting competition law into a universal compulsory-sharing regime.

Lesson

Access remedies must balance competition against incentives to invest and innovate.

24. Case 8 — Eturas v. Lietuvos Respublikos konkurencijos taryba

Case C-74/14

Importance

This European Union case is particularly relevant to automated systems because it involved an electronic platform and communications concerning discounting.

The case considered whether businesses using a common electronic system could be held responsible for coordinated conduct in circumstances where the platform facilitated the implementation of restrictions.

Relevance to machine-administered competition

The case demonstrates why digital infrastructure can be relevant to competition-law analysis.

A platform does not become legally irrelevant merely because coordination occurs through software rather than a traditional physical meeting.

Lesson

Electronic systems can facilitate conduct that competition law may scrutinize as coordinated behavior.

25. Why Eturas Is Particularly Important

The case illustrates an important transition:

Traditional model

Human meeting → agreement → cartel

Digital model

Electronic platform → communication → automated implementation → market effect

This makes digital evidence increasingly important.

Competition authorities may need to examine:

• platform messages;
• system settings;
• access permissions;
• algorithmic rules;
• user activity;
• technical logs.

26. Machine-Administered Competition Enforcement

Competition authorities may use AI in four major areas.

26.1 Market monitoring

Continuous monitoring of:

• prices;
• output;
• market shares;
• tender results.

26.2 Cartel detection

Identifying suspicious patterns.

26.3 Merger screening

Identifying potentially problematic transactions.

26.4 Digital-platform investigations

Analyzing enormous datasets involving:

• ranking;
• recommendations;
• advertising;
• consumer behavior.

27. Advantages of Machine Administration

Machine-based competition enforcement can provide:

Speed

Large datasets can be processed rapidly.

Scale

Millions of transactions can be analyzed.

Pattern detection

Algorithms can identify patterns humans may overlook.

Continuous monitoring

Markets can be monitored in real time.

Consistency

Standardized analytical processes can be applied across cases.

28. Risks of Machine Administration

However, automated competition enforcement creates risks.

28.1 False positives

Legitimate competitive behavior may appear suspicious.

28.2 False negatives

Sophisticated anticompetitive conduct may evade detection.

28.3 Algorithmic bias

Training data may contain systematic distortions.

28.4 Opacity

The regulator may not understand why a system produced a particular result.

28.5 Over-reliance

Officials may treat machine outputs as inherently correct.

28.6 Strategic manipulation

Businesses could modify their behavior to avoid detection.

29. Automated Competition Decisions

A particularly sensitive issue is whether machines should make final competition decisions.

For example:

AI system → detects suspected cartel → automatically imposes penalty

Such a system raises serious questions concerning:

• due process;
• proportionality;
• evidence;
• accountability;
• appeal rights.

A safer institutional model is:

AI → detection

Human investigators → verification

Legal authority → decision

Court → review

30. Algorithmic Regulatory Capture

Another emerging concern is algorithmic regulatory capture.

Suppose regulators rely heavily on technology supplied by dominant firms.

The regulator could become dependent upon:

• proprietary AI systems;
• proprietary datasets;
• cloud infrastructure;
• technical expertise.

This creates a risk that the regulated company indirectly controls part of the regulator's technological capability.

Competition institutions should therefore maintain:

• technological independence;
• audit capacity;
• alternative systems;
• secure infrastructure.

31. Machine Administration and Transparency

A competition framework should establish appropriate transparency standards.

Regulators should be able to determine:

1. what the algorithm does;
2. what data it uses;
3. what assumptions it makes;
4. what limitations exist;
5. how results can be reproduced.

But transparency does not necessarily require public disclosure of confidential source code.

32. Algorithmic Explainability

Explainability should be proportionate to the legal significance of the decision.

Low-risk decision

A basic explanation may be sufficient.

High-impact enforcement decision

Greater explanation may be necessary.

For example:

“The system classified this firm as high cartel risk.”

should not be the complete legal reasoning.

The authority should also be able to identify the underlying evidence and reasoning supporting further action.

33. Machine-Administered Competition and Judicial Review

Courts may increasingly encounter disputes involving:

• algorithmic evidence;
• machine-generated economic models;
• AI-assisted investigations;
• automated pricing systems.

Judicial review must remain capable of examining:

• methodology;
• evidence;
• legal standards;
• procedural fairness.

The use of sophisticated technology should not make administrative decisions immune from legal scrutiny.

34. Competition Compliance by Machines

Companies may also use AI for compliance.

An internal system could monitor:

• employee communications;
• pricing;
• competitor interactions;
• bidding behavior;
• discounts.

It could alert management:

“Potential competition-law risk detected.”

This can strengthen compliance.

However, companies remain responsible for ensuring that automated compliance systems are appropriately designed and monitored.

35. Autonomous Competition Compliance

The future may involve:

AI compliance agent

↓

monitors employee activity

↓

detects possible cartel behavior

↓

blocks risky communication

↓

alerts legal department

This could substantially improve preventive competition compliance.

However, it should not replace legal judgment in complex cases.

36. Machine-Administered Markets and Consumer Welfare

Machines can increase efficiency through:

• lower transaction costs;
• faster price comparison;
• optimized logistics;
• personalized recommendations;
• automated procurement.

But they can also reduce competition through:

• lock-in;
• coordinated pricing;
• discriminatory ranking;
• exclusionary algorithms.

Therefore, competition policy should remain technology-neutral.

The question should not be:

“Is the conduct automated?”

The better question is:

“What is the competitive effect of the automated conduct?”

37. Regulatory Framework for Machine-Administered Competition

A future framework could contain seven components.

1. Algorithmic accountability

Identify the entity responsible for deploying the system.

2. Auditability

Maintain appropriate records of significant automated decisions.

3. Human oversight

Retain human review for major enforcement decisions.

4. Competition monitoring

Continuously monitor high-risk markets.

5. Interoperability

Prevent unjustified technical exclusion.

6. Merger surveillance

Monitor acquisitions involving emerging technologies.

7. Procedural safeguards

Guarantee notice, evidence access, hearing and appeal rights.

38. Machine Administration and International Competition

AI markets frequently operate across borders.

An algorithm may be:

• designed in Country A;
• hosted in Country B;
• trained using data from Country C;
• operated by a company in Country D;
• used by consumers globally.

This creates difficult jurisdictional questions.

International cooperation may therefore be required for:

• evidence gathering;
• merger investigations;
• cartel detection;
• cross-border enforcement;
• technical standards.

39. Long-Term Evolution

Machine-administered competition frameworks may evolve through several stages.

Stage 1 — Algorithm-assisted markets

Humans make decisions with software assistance.

Stage 2 — Algorithm-managed markets

Algorithms make routine commercial decisions.

Stage 3 — AI-managed ecosystems

AI systems determine significant commercial strategies.

Stage 4 — Autonomous economic agents

Machines negotiate directly with other machines.

Stage 5 — Machine-assisted enforcement

Competition authorities rely extensively on AI for investigations.

Stage 6 — Integrated computational competition governance

Markets and enforcement systems become highly automated while legal responsibility remains with identifiable human and institutional actors.

40. Fundamental Legal Principles

A machine-administered competition framework should preserve several principles.

Principle 1 — Automation neutrality

Automation should not automatically make conduct lawful or unlawful.

Principle 2 — Accountability

An economic undertaking should not escape responsibility merely because software executed the conduct.

Principle 3 — Human oversight

Important legal decisions should remain reviewable by accountable institutions.

Principle 4 — Explainability

Affected parties should receive adequate reasons for significant decisions.

Principle 5 — Evidence integrity

Machine-generated evidence should be verifiable.

Principle 6 — Innovation protection

Regulation should not unnecessarily discourage technological investment.

Principle 7 — Competitive neutrality

Similar competitive conduct should be treated consistently regardless of whether it is performed by humans or machines.

41. Important Case-Law Table

42. Conclusion

Machine-administered competition frameworks represent a major transformation in the administration of modern competition law.

Machines can now participate in both sides of the competition equation:

Market side:
AI and algorithms determine prices, rankings, access, transactions and resource allocation.

Regulatory side:
AI can monitor markets, identify suspicious conduct, analyze mergers and assist investigations.

The central legal challenge is therefore to combine technological efficiency with legal accountability.

The cases of Microsoft, Terminal Railroad, United Brands, Google Shopping, Intel, Aspen Skiing, Trinko and Eturas demonstrate that established competition principles remain relevant even when market conduct moves into highly automated environments.

The fundamental rule should be:

Machines may administer competitive processes, but they should not become a mechanism for escaping competition-law responsibility or eliminating meaningful competitive choice.

Quick Revision Points

• Machine-administered competition = competition systems substantially operated or monitored through algorithms and AI.
• It has two dimensions: machine-administered markets and machine-assisted enforcement.
• Algorithms can determine prices, rankings, access and allocation.
• Automated systems can facilitate collusion.
• Digital platforms can become gatekeepers.
• AI ranking can create self-preferencing concerns.
• Regulators can use AI for cartel and merger screening.
• Machine-generated evidence must remain verifiable.
• A machine output should not automatically equal a legal conclusion.
• Human oversight is important for major enforcement decisions.
• Automation does not automatically remove corporate responsibility.
• Microsoft → platform power.
• Terminal Railroad → critical infrastructure.
• United Brands → dominance.
• Google Shopping → ranking and self-preferencing.
• Intel → exclusionary incentives.
• Aspen Skiing → refusal to deal.
• Trinko → limits on compulsory access.
• Eturas → electronic platform-facilitated coordination.
• The long-term objective is efficient machine-assisted competition with accountable human and institutional governance.