Competition Law and Machine-Enabled Coordination Risks

1. Introduction

Machine-enabled coordination refers to situations where algorithms, artificial intelligence (AI), machine-learning systems, automated pricing tools, or autonomous decision systems facilitate coordination between competing businesses.

Competition law traditionally addresses coordination through concepts such as:

• cartels;
• price fixing;
• market allocation;
• bid rigging;
• information exchange;
• concerted practices;
• tacit coordination;
• hub-and-spoke arrangements.

Machine-enabled markets create an additional problem because competitors may use algorithms that observe market conditions, react to competitors' behaviour, exchange information through digital systems, or independently learn strategies that produce coordinated outcomes.

The important distinction is:

Parallel behaviour or similar algorithmic decisions are not automatically unlawful. Competition law generally requires evidence of an agreement, concerted practice, or other legally prohibited conduct, depending on the jurisdiction.

2. Meaning of Machine-Enabled Coordination

Machine-enabled coordination can occur when technology assists competitors in reaching or maintaining a coordinated market outcome.

Examples include:

1. Algorithmic price coordination

Competitors use pricing algorithms that continuously monitor each other's prices.

2. Automated information exchange

Software automatically collects or communicates sensitive information such as:

• prices;
• discounts;
• inventories;
• production levels;
• bids;
• future business strategies.

3. Common pricing software

Several competitors use the same third-party algorithm to determine prices.

4. Algorithmic signalling

A system may rapidly communicate market intentions through observable price changes or other market signals.

5. Autonomous learning

Machine-learning systems may independently discover pricing strategies that reduce competitive rivalry.

6. Platform-facilitated coordination

A common digital platform may allow competing firms to interact through a central technological system.

3. Why Machine-Enabled Coordination Is Different

Traditional cartels generally involve human communication.

For example:

Company A's manager calls Company B's manager and agrees to increase prices.

Machine-enabled coordination may instead involve:

Company A and Company B independently deploy algorithms that continuously monitor each other and adjust prices according to predetermined rules.

This creates a difficult legal question:

When does technological interdependence become unlawful coordination?

The answer depends on the applicable competition-law regime and evidence.

4. Main Competition-Law Risks

4.1 Price Fixing

Price fixing is one of the most serious competition-law concerns.

Competitors may use algorithms to:

• establish prices;
• monitor competitors;
• punish deviations;
• maintain price levels;
• coordinate discounts.

The algorithm does not necessarily change the underlying legal character of the conduct.

If competitors intentionally use technology to implement an agreement to fix prices, the technology can be viewed as the means of implementing the coordination.

5. Tacit Coordination

Tacit coordination is more difficult.

It may occur when firms independently recognize that certain behaviour is mutually beneficial without an explicit agreement.

For example:

• Firm A increases price.
• Firm B's algorithm detects it.
• Firm B automatically increases price.
• Firm A's algorithm detects B's increase.
• Both maintain higher prices.

This may produce coordinated outcomes without direct communication.

However:

Similar or parallel prices alone do not automatically establish an unlawful cartel.

Authorities normally need to identify the legal basis for liability under the applicable jurisdiction.

6. Algorithmic Collusion

Algorithmic collusion can broadly be divided into four models.

Model 1: Messenger algorithm

Humans agree on the conduct, and algorithms execute it.

This is the simplest case.

Model 2: Predictable reaction

Competitors know that their algorithms will automatically react to each other's actions.

Model 3: Common intermediary

Multiple competitors use the same platform or algorithmic intermediary.

Model 4: Autonomous learning

Machine-learning systems independently discover strategies that result in coordinated behaviour.

The fourth model creates the most difficult questions concerning proof and legal responsibility.

7. Hub-and-Spoke Coordination

A platform can sometimes function as the hub connecting competing businesses.

For example:

Retailer A → Algorithm/platform ← Retailer B → Retailer C

If the platform facilitates the exchange of competitively sensitive information or coordinates pricing, competition authorities may investigate whether the arrangement constitutes an unlawful concerted practice.

The existence of a common platform alone, however, does not prove an infringement.

8. Common Algorithm Problem

Suppose competing companies independently purchase the same pricing software.

The software recommends similar prices to all users.

The result is:

Competitor A: ₹1,000
Competitor B: ₹1,000
Competitor C: ₹1,000

The identical price does not by itself establish a cartel.

Authorities would need to investigate:

• what information the software receives;
• whether competitors communicate through it;
• whether the provider facilitates coordination;
• whether firms knew how the system worked;
• whether sensitive information is shared;
• whether the arrangement intentionally reduces competition.

9. Third-Party Algorithm as a Coordination Mechanism

A particularly important scenario involves a common algorithm provider.

Suppose a third-party company provides pricing software to 500 competing businesses.

The software collects:

• competitors' prices;
• inventory;
• sales;
• demand;
• discounts.

It then recommends prices to each customer.

The competition concern is whether the system merely provides legitimate market intelligence or becomes a coordination mechanism.

10. Machine-Enabled Information Exchange

Competition law is particularly concerned with exchanges of competitively sensitive information.

Examples include:

• future prices;
• future output;
• customer allocation;
• strategic plans;
• production capacity;
• bidding intentions.

Algorithms can make information exchange:

• faster;
• broader;
• continuous;
• difficult to detect;
• more precise.

This may increase the risk of coordination.

11. Real-Time Market Monitoring

Traditional firms may monitor competitors periodically.

Algorithms can monitor them continuously.

For example:

09:00 — Competitor A changes price

09:00:01 — Algorithm detects change

09:00:02 — Competitor B's algorithm responds

09:00:03 — Competitor C responds

This creates an extremely rapid competitive feedback loop.

The legal significance depends on whether the conduct reflects:

• lawful independent adaptation; or
• unlawful agreement/concerted practice.

12. Bid-Rigging Algorithms

Machine systems can also affect public procurement.

Possible risks include:

• automated bid rotation;
• allocation of contracts;
• predetermined bidding ranges;
• coordinated withdrawal;
• automatic undercutting;
• exchange of tender information.

Because procurement markets often involve repeated transactions, algorithms may make coordinated strategies easier to implement.

13. Market Allocation

Algorithms can potentially facilitate allocation of:

• customers;
• geographic territories;
• products;
• delivery areas;
• online advertising opportunities.

For example:

Algorithm A serves customers in Northern India while Algorithm B avoids those customers.

If this reflects an agreement between competitors, it may raise serious market-allocation concerns.

14. Output Coordination

Coordination does not have to involve prices.

Algorithms could potentially coordinate:

• production;
• inventory;
• supply;
• capacity;
• advertising;
• distribution.

Reduced output can produce higher prices even when no explicit price agreement exists.

15. Machine Learning and Adaptive Coordination

Machine-learning systems are different from simple rules-based algorithms.

A traditional algorithm might state:

“If competitor price increases by 5%, increase our price by 5%.”

A machine-learning system might instead learn:

“When competitors behave in a particular way, a certain pricing strategy maximizes long-term revenue.”

The system could modify its behaviour over time.

This creates uncertainty about:

• explainability;
• foreseeability;
• accountability;
• intention;
• evidence;
• causation.

16. Autonomous Algorithmic Coordination

The most difficult hypothetical scenario is:

Two competing firms deploy autonomous AI systems that independently learn that maintaining higher prices maximizes long-term profits.

No employee communicates with the rival.

The machines nevertheless develop coordinated behaviour.

This creates a major conceptual distinction:

Economic coordination

The market outcome becomes coordinated.

Legal coordination

The conduct satisfies the legal requirements for an infringement.

The two concepts should not automatically be treated as identical.

17. Case Law

Because machine-enabled coordination is a developing area, most important cases are foundational or analogous authorities rather than cases involving modern autonomous AI systems directly.

Case 1: United States v. Socony-Vacuum Oil Co.

Citation: 310 U.S. 150 (1940)

Principle

The U.S. Supreme Court treated agreements among competitors to influence or fix prices as a core Sherman Act violation.

Relevance

If competitors use algorithms as the mechanism for implementing an underlying agreement to fix prices, the use of technology does not make the underlying arrangement lawful.

18. Case 2: Interstate Circuit, Inc. v. United States

Citation: 306 U.S. 208 (1939)

Principle

The Court recognized that an agreement or concerted arrangement may sometimes be inferred from surrounding circumstances rather than from a single written contract.

Relevance

This is important for algorithmic coordination because digital coordination may leave fewer traditional communications.

Authorities may examine:

• communications;
• system design;
• data flows;
• pricing behaviour;
• contractual arrangements;
• knowledge;
• surrounding circumstances.

The case therefore illustrates the importance of circumstantial evidence.

19. Case 3: American Tobacco Co. v. United States

Citation: 328 U.S. 781 (1946)

Principle

The Supreme Court recognized that unlawful concerted conduct may be established through a combination of circumstances and conduct.

Relevance

Machine-enabled coordination may similarly require authorities to reconstruct the surrounding technological and commercial environment rather than search only for an explicit human conversation.

20. Case 4: A. Ahlström Osakeyhtiö and Others v. Commission — Wood Pulp

Citation: Joined Cases 89/85 and others, European Court of Justice, 1988

Principle

The case concerned parallel behaviour and the evidentiary distinction between conscious parallelism and concerted practices.

Relevance

It is particularly useful for algorithmic markets.

Algorithms frequently react to publicly observable market information.

Therefore:

Parallel algorithmic pricing is not automatically proof of collusion.

Authorities must distinguish lawful independent adaptation from coordination involving the necessary legal elements of a concerted practice.

21. Case 5: Eturas UAB and Others

Citation: Case C-74/14, Court of Justice of the European Union, 2016

Facts

The case concerned an online booking platform where a technological system communicated a restriction affecting discounts offered by participating travel agencies.

Principle

The Court examined whether knowledge of a platform-based restriction and continued participation could support liability for concerted practice.

Relevance

This is one of the most directly relevant authorities for digital coordination.

It demonstrates that:

• technology can facilitate coordination;
• electronic communications can be relevant evidence;
• participation in a digital system can have competition-law consequences;
• knowledge and conduct of participants matter.

22. Case 6: AC-Treuhand v. Commission

Citation: Case C-194/14 P, Court of Justice of the European Union, 2015

Principle

The Court confirmed that a non-competitor undertaking can, in appropriate circumstances, be liable for facilitating a cartel.

Relevance

This is highly relevant to third-party algorithm providers.

Suppose:

Competitor A + Competitor B + Algorithm Provider

The algorithm provider may not compete in the downstream market.

Nevertheless, if it knowingly facilitates unlawful coordination, its role may become competition-law relevant.

23. Case 7: Groupement des Cartes Bancaires v. Commission

Citation: Case C-67/13 P, Court of Justice of the European Union, 2014

Principle

The Court emphasized that conduct cannot simply be classified as a restriction “by object” without proper examination of its content, objectives, and legal and economic context.

Relevance

This is important for AI-based enforcement.

A regulator should not conclude:

“Algorithms produced similar prices → therefore cartel.”

The legal classification must be supported by appropriate analysis.

This helps prevent false positives in algorithmic cartel detection.

24. Case 8: Piau v. Commission

Citation: Case T-193/02, General Court, 2005

Principle

The case concerned rules governing football agents and competition-law issues arising from regulatory arrangements.

Relevance

It provides a broader illustration of how coordinated rules and intermediary structures may require examination of their effects on competition.

For machine-enabled markets, intermediary platforms can similarly become important in analysing coordination.

25. Case 9: European Commission v. Anheuser-Busch / Eturas Line of Digital Cases

The wider European digital-platform enforcement experience demonstrates that competition authorities increasingly examine electronic systems as possible mechanisms through which competitors can coordinate.

The important legal lesson is not that technology itself creates liability.

Rather:

Digital architecture can constitute evidence of how coordination occurred and who participated in it.

26. Case 10: United States v. Apple Inc.

U.S. e-books antitrust litigation

The Apple e-books litigation is relevant by analogy because it demonstrated how coordination can occur through a platform and intermediary contractual structure rather than through a simple traditional cartel meeting.

Relevance to machine-enabled markets

Modern digital systems can similarly create:

• common pricing mechanisms;
• common contractual structures;
• information flows;
• intermediary coordination.

The legal analysis must focus on the actual arrangement and evidence.

27. Algorithmic Coordination vs. Independent Parallel Conduct

This distinction is fundamental.

28. Evidence in Machine-Enabled Coordination Cases

Traditional cartel investigations may rely on:

• emails;
• meetings;
• telephone records;
• contracts;
• witness testimony.

Algorithmic investigations may additionally require:

Technical evidence

• source code;
• model architecture;
• APIs;
• logs;
• system instructions;
• training data;
• model outputs.

Commercial evidence

• pricing strategies;
• contracts;
• internal policies;
• communications;
• market data.

Economic evidence

• price movements;
• margins;
• output;
• market shares;
• demand patterns.

29. Importance of Algorithm Logs

Algorithm logs can show:

• when an algorithm changed strategy;
• which information it received;
• which competitors it monitored;
• what output it generated;
• whether humans overrode its decisions.

These records may become important evidence in enforcement proceedings.

30. Explainability

Competition authorities may need to understand:

Why did the algorithm increase prices?

Possible explanations include:

1. increased demand;
2. increased input costs;
3. competitor price changes;
4. deliberate coordination;
5. machine-learning adaptation.

An unexplained algorithmic output should therefore not automatically be treated as proof of collusion.

31. Human Responsibility

A company cannot ordinarily avoid competition-law obligations simply by saying:

“The algorithm did it.”

Human decision-makers may remain responsible where they:

• designed the system;
• instructed the system;
• knowingly used coordinated outputs;
• ignored obvious competition risks;
• intentionally created a mechanism for coordination.

32. Third-Party Technology Providers

Competition law may also become relevant to:

• pricing-software companies;
• marketplace operators;
• data providers;
• AI developers;
• consulting firms;
• cloud platforms.

A provider that merely supplies neutral software is not automatically liable.

The legal question is whether it knowingly and materially participates in or facilitates unlawful coordination.

33. Machine-Enabled Coordination in Digital Platforms

Platforms can create particularly strong coordination risks because they may control:

• prices;
• rankings;
• product visibility;
• customer data;
• transaction information;
• recommendations;
• seller communications.

The platform's role should therefore be examined carefully where competitors use a common digital environment.

34. Algorithmic Monitoring by Competition Authorities

Competition authorities can themselves use machine learning to detect possible cartels.

They may identify:

• unusual price synchronization;
• bid rotation;
• identical discounts;
• suspicious bidding patterns;
• geographic allocation;
• sudden parallel price changes.

But such systems should generally be treated as screening and investigative tools, not automatic proof of infringement.

35. False Positives

A major problem is that legitimate markets can display highly similar prices.

For example:

If three companies face the same increase in raw-material costs, their algorithms may independently increase prices.

An AI detection system could identify:

“High probability of coordination.”

But economic correlation does not necessarily establish an unlawful agreement.

Human/legal investigation remains necessary.

36. False Negatives

The opposite problem also exists.

A sophisticated cartel could intentionally design algorithms to:

• disguise communication;
• avoid identical prices;
• introduce random variations;
• conceal market allocation;
• make coordination appear independent.

Therefore, competition authorities cannot rely exclusively on simple statistical tests.

37. Compliance Measures for Businesses

Companies using AI pricing systems should establish:

1. Competition-law policies

Employees should understand what information cannot be exchanged.

2. Algorithm audits

Regularly test whether systems create coordination risks.

3. Data controls

Restrict access to competitors' sensitive information.

4. Human oversight

Require review of high-risk pricing decisions.

5. Documentation

Maintain records of:

• system design;
• objectives;
• data sources;
• safeguards.

6. Third-party review

Review pricing software supplied by external vendors.

38. Regulatory Approach

Competition authorities can adopt a combination of:

Ex-ante measures

• compliance guidance;
• algorithm audits;
• information-exchange safeguards;
• platform rules.

Ex-post enforcement

• cartel investigations;
• dawn raids;
• forensic analysis;
• economic analysis;
• penalties;
• behavioural remedies.

39. Key Legal Principles

The following principles are particularly important:

1. Algorithms are tools; legal responsibility generally remains with undertakings and relevant participants.
2. Parallel algorithmic behaviour is not automatically a cartel.
3. An explicit agreement implemented through algorithms remains potentially unlawful.
4. Third-party facilitators may face liability in appropriate circumstances.
5. Sensitive information exchange creates significant competition risks.
6. Common software does not automatically establish collusion.
7. Economic evidence should be interpreted in its legal context.
8. AI detection should generate investigative leads, not automatically determine liability.
9. Human oversight is important for high-risk algorithmic systems.
10. The distinction between economic coordination and legally prohibited coordination must be maintained.

40. Revision Table

41. Conclusion

Machine-enabled coordination creates a new technological dimension to traditional competition-law problems, but it does not eliminate the traditional legal requirements for establishing an infringement.

The most important risks arise where algorithms:

• implement an existing cartel;
• facilitate exchange of competitively sensitive information;
• enable hub-and-spoke coordination;
• monitor competitors in a manner designed to sustain coordination;
• automate bid rigging or market allocation;
• are knowingly designed or used to facilitate unlawful coordination.

The cases of Socony-Vacuum, Interstate Circuit, American Tobacco, Wood Pulp, Eturas, AC-Treuhand, Groupement des Cartes Bancaires, and the Apple e-books litigation provide important foundations for analysing these issues.

The central principle is:

Technology may change how coordination occurs, but it does not by itself determine whether the coordination is legally unlawful. Competition authorities must distinguish legitimate algorithmic adaptation and parallel conduct from conduct supported by an agreement, concerted practice, or other legally sufficient basis for liability.