Competition Law and AI-Assisted Antitrust Enforcement

1. Introduction

AI-assisted antitrust enforcement means the use of artificial intelligence, machine learning, data analytics, automated screening and algorithmic tools by competition authorities, courts, businesses and investigators to identify possible violations of competition law.

AI can assist enforcement by examining very large datasets that would be difficult to analyse manually. For example, an authority may use algorithms to identify:

• unusual price movements;
• parallel pricing between competitors;
• suspicious bidding patterns;
• possible bid rotation;
• market-sharing patterns;
• coordinated conduct through digital platforms;
• discriminatory algorithms;
• exclusionary conduct by dominant firms;
• suspicious mergers or acquisitions;
• exchange of commercially sensitive information;
• algorithmic price-fixing arrangements.

The important legal point is that AI is generally an investigative or analytical tool; the underlying competition-law violation still has to be established under the applicable legal standard. In EU law, for example, Article 101 TFEU addresses anti-competitive agreements and concerted practices, while Article 102 addresses abusive conduct by dominant undertakings.

2. What Is AI-Assisted Antitrust Enforcement?

AI-assisted enforcement can be divided into two different situations.

A. AI used by enforcement authorities

A competition authority may use AI to:

1. screen millions of transactions;
2. identify suspicious price patterns;
3. detect possible cartel communications;
4. analyse procurement bids;
5. identify relationships between companies;
6. examine large quantities of emails and documents;
7. identify potentially relevant evidence;
8. prioritise investigations.

Here, AI assists the enforcement process.

B. AI used by companies

Companies may themselves use AI or pricing algorithms to determine:

• prices;
• discounts;
• inventory;
• bidding strategies;
• customer targeting;
• output levels.

This creates another competition-law problem when algorithms are used to facilitate coordination between competitors.

The second issue has already generated important litigation, including United States v. Topkins and the more recent RealPage litigation.

3. Why AI Creates New Antitrust Enforcement Problems

Traditional competition investigations often depend upon evidence such as:

• emails;
• contracts;
• telephone calls;
• meetings;
• pricing documents;
• internal company communications.

AI-assisted markets can make the evidence much more complicated.

For example, two competitors might never communicate directly. Instead, both may use the same algorithmic pricing system.

The legal question becomes:

When does algorithmic similarity represent legitimate independent conduct, and when does it represent unlawful coordination?

This is one of the central unresolved questions in algorithmic antitrust enforcement. Legal scholarship continues to distinguish between explicit algorithmic collusion, hub-and-spoke coordination and potentially autonomous algorithmic coordination.

4. Major Areas of AI-Assisted Antitrust Enforcement

A. Algorithmic price-fixing

Suppose five competitors use AI systems to determine their prices.

If the systems independently produce similar prices, that fact alone does not automatically prove an unlawful agreement.

However, if competitors:

• exchange competitively sensitive information;
• agree to use the same pricing mechanism;
• instruct an algorithm to maintain agreed prices; or
• use a common intermediary to coordinate pricing,

competition law may become applicable.

The important distinction is between parallel conduct and concerted conduct.

B. AI-powered cartel detection

Competition authorities can use machine learning to identify patterns that may indicate a cartel.

For example:

However, these indicators should generally be treated as investigative leads rather than automatic proof of infringement.

A statistical correlation may have legitimate explanations.

5. AI and the Definition of the Relevant Market

AI may also assist authorities in defining markets.

Traditional market definition examines:

• substitutability;
• consumer preferences;
• prices;
• geographic scope;
• product characteristics.

AI can analyse:

• millions of consumer transactions;
• search behaviour;
• switching patterns;
• online purchasing data;
• customer reviews;
• pricing histories.

This may be particularly important in digital markets where products can be supplied for zero monetary prices or where competition takes place through:

• data;
• quality;
• innovation;
• privacy;
• interoperability;
• ecosystem access.

6. AI and Abuse of Dominance

AI can also help authorities investigate dominant companies.

Potential concerns include:

Self-preferencing

An AI-driven platform could systematically favour its own products.

Discriminatory ranking

Algorithms could give competitors less favourable rankings.

Predatory pricing

AI could be used to optimise extremely low prices for particular customers or markets.

Exclusive dealing

AI systems could identify and target customers in ways that strengthen exclusionary strategies.

Refusal of access

A dominant platform could use automated systems to determine which competitors receive access to important infrastructure or data.

EU Article 102 enforcement focuses on whether a dominant undertaking has engaged in abusive exclusionary conduct, with market definition and dominance forming central parts of the analysis.

7. AI and Merger Enforcement

AI can assist merger investigations by examining:

• ownership structures;
• acquisition histories;
• patent portfolios;
• technology overlaps;
• customer data;
• product markets;
• competitor relationships;
• investment patterns.

This is particularly important for technology markets because an acquisition of a relatively small company may nevertheless provide a major platform with valuable:

• AI talent;
• data;
• intellectual property;
• algorithms;
• infrastructure;
• emerging technology.

AI therefore has the potential to help authorities identify competitive relationships that may be difficult to discover through traditional document review.

8. Important Case Laws

There is not yet a large body of reported judgments specifically deciding the legality of competition authorities using AI themselves for enforcement. Therefore, the most useful case law consists of cases establishing principles for algorithmic coordination, digital-platform conduct, information exchange and evidence—principles that can be applied when AI is involved.

Case 1 — United States v. Topkins

Jurisdiction: United States
Court: U.S. District Court, Northern District of California
Year: 2015

Facts

David Topkins and other online sellers agreed to coordinate prices for posters sold through an online marketplace.

The competitors did not merely discuss prices. They agreed to use pricing algorithms to implement their price-fixing arrangement.

Topkins pleaded guilty.

Legal principle

The use of software does not transform an unlawful price-fixing agreement into lawful conduct.

The important point is that the human agreement existed first, while the algorithm was used to implement the agreement.

Importance for AI enforcement

This case establishes an important starting point:

If competitors agree to fix prices, using AI or an algorithm to execute the agreement does not remove antitrust liability.

AI can therefore be treated as the technological mechanism through which a conventional cartel operates.

9. Case 2 — Eturas v. Lietuvos Respublikos konkurencijos taryba

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

Facts

Travel agencies used the E-TURAS online booking system.

A system message communicated a limitation concerning discounts that could be offered to customers.

The European competition authorities considered whether participating businesses could be treated as having participated in a concerted practice.

Legal principle

The existence of a message distributed through a common technological platform does not automatically establish that every participant participated in an anti-competitive agreement.

However, knowledge of the anti-competitive measure, together with circumstances demonstrating participation or failure to distance oneself where relevant, can be important.

Importance for AI

This case is extremely relevant to AI-mediated markets.

A common AI system might communicate or implement a pricing rule affecting multiple competitors.

The authority must still establish the required elements of concerted conduct.

Therefore:

Common algorithm ≠ automatic cartel.

The surrounding evidence matters.

10. Case 3 — Meyer v. Kalanick

Court: U.S. District Court for the Southern District of New York
Year: 2016

Facts

Uber drivers used Uber's pricing system.

The plaintiff alleged that Uber's algorithm facilitated coordination among drivers by preventing ordinary price competition between them.

Legal significance

The court allowed the antitrust claim to proceed at the pleading stage.

The case raised the question whether a platform's pricing mechanism could facilitate an agreement among otherwise competing service providers.

Importance for AI-assisted enforcement

The case illustrates the hub-and-spoke problem.

The structure can be represented as:

Competitor A → AI/platform → Competitor B

Instead of competitors communicating directly, a platform or algorithm can potentially become the coordinating mechanism.

This is particularly important for AI systems that collect information from competing businesses and generate common pricing recommendations.

11. Case 4 — United States v. RealPage, Inc.

Jurisdiction: United States
Proceedings: Federal antitrust litigation
Period: 2024 onward

Facts

The U.S. Department of Justice and state authorities brought proceedings concerning RealPage's rental-pricing software.

The allegations concerned the use of competitively sensitive information and algorithmic pricing recommendations involving rental housing.

The central theory involved competitors using a common pricing system that could incorporate non-public information.

Legal importance

The case demonstrates how traditional antitrust concepts can be applied to sophisticated algorithmic pricing systems.

The central issues include:

• exchange of competitively sensitive information;
• use of common algorithms;
• coordination through an intermediary;
• pricing recommendations;
• the existence of an agreement or concerted action.

The DOJ's case has been described as part of the growing antitrust focus on algorithmic pricing.

Importance for AI

RealPage illustrates why authorities increasingly examine the data entering an algorithm, not merely the output produced by it.

An apparently independent AI recommendation may require closer scrutiny if its underlying data is supplied by competing firms.

12. Case 5 — United States v. Yardi Systems / Algorithmic Pricing Litigation

Jurisdiction: United States

Yardi-related litigation concerns allegations surrounding the use of algorithmic pricing systems in rental housing.

The broader litigation raises an important question:

Can competing businesses coordinate indirectly by delegating pricing decisions to a common algorithmic intermediary?

Competition-law significance

Traditional antitrust law normally looks for:

1. competitors;
2. an agreement or concerted practice;
3. anti-competitive conduct;
4. competitive harm.

Algorithmic systems can complicate step 2 because the competitors may communicate indirectly through software.

Importance

The case illustrates the increasing importance of examining:

• what information competitors provide to the algorithm;
• whether competitors know how the system operates;
• whether the algorithm restricts independent decision-making;
• whether competitors consciously participate in the same pricing mechanism.

This area remains legally contested rather than fully settled.

13. Case 6 — Trod Ltd / GB Eye Ltd

Jurisdiction: United Kingdom
Authority: Competition and Markets Authority

Facts

Online sellers of posters and frames used automated repricing software.

The businesses had agreed not to undercut each other for certain products.

The automated software then helped maintain the agreed pricing.

Legal principle

Automated software does not provide immunity from competition law.

The CMA treated the underlying agreement between competitors as the critical issue.

Importance for AI

The case demonstrates a basic but important principle:

Automation cannot legalise an agreement that would otherwise violate competition law.

This is directly relevant to AI-assisted pricing systems.

14. Case 7 — Samir Agarwal v. ANI Technologies / Ola

Jurisdiction: India
Authority: Competition Commission of India and subsequent appellate proceedings

Facts

The case concerned allegations relating to pricing and coordination in the radio-taxi market.

The allegations raised questions about whether the platform's pricing system could facilitate coordination among competing drivers.

Competition-law significance

The case is important in India because it demonstrates the difficulty of proving a cartel or hub-and-spoke arrangement in a platform economy.

The existence of a common technological platform or pricing mechanism is not, by itself, sufficient to establish every element of a cartel.

Relevance to AI

The same analytical problem arises with AI:

Shared algorithm + similar prices ≠ automatically an unlawful agreement.

The authority must establish the legally relevant form of coordination.

15. What These Cases Show Collectively

The cases demonstrate several important principles.

16. AI as an Investigative Tool

Competition authorities can use AI in several stages.

Stage 1 — Screening

AI scans large datasets for unusual patterns.

Stage 2 — Risk identification

The system identifies transactions or companies requiring closer examination.

Stage 3 — Evidence review

Natural-language-processing systems can help organise:

• emails;
• contracts;
• internal messages;
• pricing records;
• meeting documents.

Stage 4 — Economic analysis

Machine-learning models can analyse:

• price movements;
• market shares;
• bidding patterns;
• customer switching;
• output levels.

Stage 5 — Human investigation

Investigators examine the relevant evidence and determine whether there is a legally sufficient case.

The distinction is important because statistical suspicion and legal proof are not the same thing.

17. Evidentiary Problems

AI-assisted enforcement raises several evidentiary questions.

A. Explainability

If an AI system identifies a suspected cartel, investigators need to understand why.

A competition authority should be able to distinguish:

“The algorithm says this is suspicious”

from

“The evidence establishes an infringement.”

B. False positives

AI can identify legitimate conduct as suspicious.

For example, competitors may independently increase prices because:

• raw-material costs increased;
• demand increased;
• taxes changed;
• supply decreased;
• a regulatory change occurred.

Therefore, parallel pricing alone may not prove collusion.

C. False negatives

AI may also fail to detect sophisticated coordination.

Competitors may deliberately structure communications to avoid obvious patterns.

Consequently, AI should not necessarily replace traditional investigative techniques.

18. Confidential Business Information

AI-assisted enforcement can involve extremely sensitive information.

Competition authorities may handle:

• pricing information;
• customer lists;
• strategic plans;
• production data;
• costs;
• future prices;
• trade secrets.

An AI system processing this material therefore requires strong safeguards concerning:

• access control;
• confidentiality;
• data retention;
• cybersecurity;
• audit trails;
• model governance.

19. Due Process and Procedural Fairness

A major legal issue is whether an investigated company should be able to understand the basis of an AI-assisted enforcement decision.

Important questions include:

1. What data was examined?
2. Was the data accurate?
3. What algorithm was used?
4. What assumptions did the model make?
5. Was the model independently validated?
6. Were alternative explanations considered?
7. Can the company challenge the evidence?
8. Was a human investigator involved in the final decision?

These issues become particularly important when AI is used to prioritise investigations or interpret large bodies of evidence.

20. AI and Competition Authority Resources

AI may significantly improve enforcement efficiency.

Traditional document review can require investigators to examine enormous quantities of material.

AI can potentially:

• classify documents;
• identify relevant communications;
• identify relationships;
• detect patterns;
• compare pricing data;
• identify suspicious transactions.

The European Commission maintains a large searchable competition-case database covering antitrust, cartels, mergers and other competition matters, illustrating the scale of information that modern competition enforcement may involve.

21. The Risk of Algorithmic Bias

AI systems can also create enforcement bias.

For example, an algorithm trained primarily on historical cartel cases might disproportionately identify conduct resembling those older cases while failing to recognise new forms of anti-competitive behaviour.

Therefore, authorities should consider:

• training-data quality;
• model accuracy;
• bias testing;
• regular validation;
• human oversight;
• transparency;
• reproducibility.

22. AI and Article 101 TFEU

Under EU competition law, Article 101 is particularly relevant to algorithmic coordination.

Potentially problematic conduct includes:

• price fixing;
• market sharing;
• output restrictions;
• exchange of sensitive information;
• coordinated pricing through algorithms.

The key issue remains whether the necessary agreement or concerted practice can be established.

An algorithm itself does not automatically create an Article 101 infringement.

23. AI and Article 102 TFEU

Article 102 becomes particularly relevant when AI is used by a dominant undertaking.

Potential conduct includes:

Algorithmic self-preferencing

The platform's own products receive preferential treatment.

Algorithmic exclusion

Competitors receive inferior access or visibility.

Data foreclosure

A dominant company restricts competitors' access to essential data.

Discriminatory algorithms

Different competitors receive materially different conditions without objective justification.

Predatory or exclusionary pricing

AI is used to implement a pricing strategy designed to exclude competitors.

The European Commission's current Article 102 framework focuses on exclusionary conduct by dominant undertakings.

24. AI and Indian Competition Law

In India, Section 3 of the Competition Act, 2002 is particularly relevant to agreements that cause or are likely to cause an appreciable adverse effect on competition.

Section 4 is relevant to abuse of dominant position.

AI therefore does not necessarily require a completely separate competition-law regime.

Existing concepts can potentially apply to:

• algorithmic price fixing;
• information exchange;
• hub-and-spoke arrangements;
• platform discrimination;
• self-preferencing;
• exclusionary conduct.

The major difficulty is adapting existing legal concepts to technology where decision-making may be partially or substantially automated.

25. AI-Assisted Enforcement vs AI-Driven Collusion

These concepts should not be confused.

This distinction is essential when studying AI and competition law.

26. Major Legal Challenges

1. Establishing an agreement

Algorithms can make coordination less visible.

2. Distinguishing parallel conduct

Similar prices do not automatically prove collusion.

3. Identifying responsibility

If an AI system makes the decision, investigators must determine the role of the human users and companies.

4. Explainability

Authorities need reliable reasons for AI-generated investigative conclusions.

5. Evidence

AI-generated results must be connected to admissible and reliable evidence.

6. Confidentiality

Competition investigations frequently involve commercially sensitive data.

7. Cross-border enforcement

AI platforms frequently operate across several jurisdictions.

8. Rapid technological change

Competition authorities must continuously update their investigative capabilities.

27. Future Development of AI-Assisted Antitrust Enforcement

The likely development of competition enforcement is toward a combination of:

AI screening + economic analysis + traditional investigation + human legal judgment.

AI is particularly useful for finding patterns that investigators might otherwise miss.

However, the legal determination should remain based on applicable competition-law standards rather than on an AI system's output alone.

Recent competition-policy developments also show increasing institutional attention to digital and AI markets. The European Commission has published work specifically addressing competition in generative AI and virtual worlds, while its competition framework continues to evolve around digital markets.

28. Conclusion

AI-assisted antitrust enforcement represents an important development in modern competition law.

The technology can help competition authorities:

• detect potential cartels;
• analyse enormous datasets;
• identify suspicious bidding;
• investigate algorithmic pricing;
• examine digital-platform behaviour;
• analyse mergers;
• identify potential exclusionary conduct.

At the same time, AI does not replace the legal requirements of competition law.

The central principles emerging from cases such as Topkins, Eturas, Meyer v. Kalanick, RealPage-related litigation, Trod/GB Eye and Samir Agarwal are that authorities must distinguish genuine independent conduct from coordinated conduct and must examine the evidence surrounding the algorithm rather than treating algorithmic similarity as automatic proof of infringement.