Competition Law and AI-Assisted Antitrust Investigations

1. Introduction

Artificial intelligence is increasingly relevant to competition-law enforcement. Competition authorities can use AI, machine learning, data analytics, natural-language processing, and automated screening to identify possible cartels, exclusionary conduct, discriminatory practices, and patterns of coordination.

At the same time, AI can itself create competition concerns. For example, competing firms may use a common pricing algorithm, dominant AI companies may control important inputs such as computing capacity or data, or an algorithm may systematically disadvantage rivals.

The important legal principle is that AI does not create a separate exemption from competition law. Existing competition rules can apply when unlawful conduct is implemented or facilitated through software or AI. The U.S. DOJ and FTC have expressly stated that companies cannot avoid antitrust law merely because conduct is carried out through algorithms.

The OECD similarly identifies algorithmic information sharing, price coordination, discrimination, exclusion, and self-preferencing as emerging competition concerns.

2. Meaning of AI-Assisted Antitrust Investigation

An AI-assisted antitrust investigation is an investigation in which competition authorities use AI or advanced computational techniques to help:

• identify suspicious market behaviour;
• analyse large quantities of business documents;
• detect communication patterns;
• examine pricing data;
• identify possible coordination between competitors;
• analyse algorithmic pricing systems;
• identify exclusionary conduct;
• reconstruct market behaviour;
• detect unusual changes in prices or output;
• examine relationships between companies;
• prioritise evidence for human investigators.

AI should generally be understood as an investigative assistance tool, rather than as an independent decision-maker.

The European Commission already has powers to conduct ex-officio investigations using market intelligence, information requests, inspections and other investigative powers.

3. Why AI Is Important for Antitrust Enforcement

Traditional investigations can involve enormous amounts of information:

• emails;
• contracts;
• internal presentations;
• WhatsApp or other communications;
• pricing databases;
• transaction records;
• customer information;
• source code;
• algorithm documentation;
• API records;
• server logs;
• business plans;
• internal financial models.

In large digital investigations, manually reviewing all of this material can be extremely difficult.

AI can assist investigators by identifying:

Relevant documents → Relevant communications → Relevant transactions → Suspicious patterns → Human investigation

The final legal conclusion should still be based on legally admissible and properly evaluated evidence.

4. Main Areas Where AI Can Assist Competition Authorities

A. Cartel Detection

AI can analyse prices across competitors and identify unusual parallel movements.

For example:

Company A: ₹100 → ₹110
Company B: ₹98 → ₹109
Company C: ₹101 → ₹111

Repeated, highly unusual parallel movements may justify further investigation.

However, parallel pricing alone does not automatically establish a cartel. Similar prices may result from common costs, demand conditions, or legitimate competitive behaviour.

This distinction is particularly important when AI produces an initial "red flag."

B. Detection of Algorithmic Collusion

AI can help identify situations where algorithms may facilitate coordination.

Potential warning signs include:

• repeated synchronised price changes;
• identical pricing responses;
• common algorithmic parameters;
• competitors using the same pricing provider;
• competitors supplying sensitive data to a common intermediary;
• algorithms reacting rapidly to competitors' prices;
• restrictions against lowering prices;
• unusual stability in competitor margins.

The legal question remains whether the conduct satisfies the relevant requirements for an agreement, concerted practice, abuse of dominance, or another infringement.

5. AI and Market Definition

Market definition is one of the first stages of many antitrust investigations.

AI can help authorities analyse:

• consumer search behaviour;
• product characteristics;
• transaction data;
• switching patterns;
• pricing;
• customer surveys;
• online behaviour;
• substitution patterns.

For example, an investigation might ask whether:

AI writing software + traditional writing software

belong to the same relevant product market.

AI tools can analyse large quantities of consumer and transaction information, but the legal market-definition exercise remains a matter of competition-law analysis.

6. AI and Detection of Dominance

AI can help investigators analyse possible dominance by examining:

• market shares;
• entry barriers;
• customer switching;
• network effects;
• access to data;
• cloud infrastructure;
• computing resources;
• distribution channels;
• interoperability;
• ecosystem dependencies.

This is particularly important in AI markets because competition may occur at several layers:

1. semiconductor hardware;
2. cloud computing;
3. computing capacity;
4. foundation models;
5. AI applications;
6. distribution platforms;
7. data;
8. AI-enabled downstream services.

The OECD has identified structural competition risks in several layers of the AI value chain, including concentrated hardware and data markets.

7. AI and Abuse of Dominance

AI can assist authorities in detecting conduct such as:

• self-preferencing;
• discriminatory access;
• tying;
• bundling;
• refusal to deal;
• exclusionary contracts;
• discriminatory ranking;
• predatory pricing;
• loyalty restrictions;
• interoperability restrictions.

For an Article 102 TFEU investigation, for example, the European Commission first considers the relevant product and geographic market and then assesses whether the undertaking holds a dominant position.

8. AI and Evidence Discovery

One of the most important applications is document discovery.

Natural-language-processing systems can classify documents according to subjects such as:

• "pricing strategy";
• "competitor information";
• "market sharing";
• "customer allocation";
• "algorithm";
• "exclusive agreement";
• "data access";
• "internal pricing."

This can reduce the volume of material investigators need to review manually.

However, AI classification can produce:

• false positives;
• false negatives;
• contextual errors;
• language errors;
• misleading correlations.

Therefore, human verification remains important.

9. AI-Assisted Pricing Analysis

Authorities can compare millions of pricing observations.

They may investigate:

• price movements;
• price dispersion;
• discounts;
• geographic pricing;
• competitor responses;
• algorithmic recommendations;
• changes before and after an agreement;
• customer-specific pricing.

The purpose is not simply to ask:

"Are the prices similar?"

Instead, investigators may ask:

"What explains the pricing pattern, and is there evidence of unlawful coordination or exclusion?"

10. AI and Competitively Sensitive Information

A particularly important issue arises when several competitors provide sensitive information to the same AI or algorithmic platform.

Potentially sensitive information includes:

• current prices;
• future prices;
• production plans;
• capacity;
• margins;
• discounts;
• customer information;
• inventory;
• strategic plans.

If a common algorithm receives such information from competing firms and uses it to generate recommendations, investigators may examine whether the arrangement reduces independent competitive decision-making.

The DOJ's RealPage enforcement is an important modern example.

11. Case Law and Enforcement Examples

Case 1 — United States v. David Topkins

Jurisdiction: United States
Year: 2015
Issue: Algorithmic price fixing

This is one of the earliest major U.S. cases involving algorithmic pricing.

David Topkins and co-conspirators sold posters through an online marketplace. According to the DOJ, the participants agreed to fix prices and implemented their agreement through pricing algorithms.

The important principle is that:

A cartel does not become lawful merely because computers implement the agreement.

Relevance to AI-assisted investigations

Investigators examining AI-enabled pricing systems may therefore look beyond the software itself and investigate:

• who designed the algorithm;
• what instructions were given;
• what information was exchanged;
• whether competitors communicated;
• whether the algorithm implemented an agreement.

12. Case 2 — Eturas UAB and Others v. Lithuanian Competition Council

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

This case concerned travel agencies using a common computerised booking system.

The system automatically restricted discounts available to customers. The case examined whether the use of a common computer system and communications through that system could constitute evidence of a concerted practice.

Importance

The case demonstrates that competition law can examine conduct implemented through a technological system.

It is particularly useful for AI investigations because investigators may need to determine:

• what the software actually did;
• what users knew;
• whether users received information from the system;
• whether they accepted or participated in the resulting conduct.

Technology does not eliminate the requirement to establish the necessary legal elements of coordination.

13. Case 3 — Samir Agrawal v. Competition Commission of India

Court: National Company Law Appellate Tribunal, India
Year: 2020
Subject: Ola/Uber algorithmic pricing

The case involved allegations that cab aggregators' algorithms facilitated price fixing between drivers.

The CCI had initially found no prima facie agreement or arrangement sufficient to establish the alleged infringement and closed the matter. The appeal challenged that conclusion.

Importance

This case is particularly relevant to AI-assisted antitrust investigations in India because it demonstrates an important evidentiary issue:

Algorithmic pricing ≠ automatically unlawful coordination.

An authority must examine whether the evidence establishes the legally required agreement, understanding, or concerted practice.

AI-generated evidence therefore has to be connected to the legal elements of Section 3 of the Competition Act.

14. Case 4 — U.S. v. RealPage

Jurisdiction: United States
Subject: Algorithmic rental pricing
Major enforcement: 2024–2026

The DOJ alleged that RealPage's pricing software used non-public, competitively sensitive information supplied by competing landlords and that the software contributed to alignment of rental pricing. The DOJ brought claims under Sections 1 and 2 of the Sherman Act.

The subsequent enforcement process resulted in settlements and proposed/final judgments involving RealPage and participating landlords. The DOJ's case materials show continuing proceedings and judgments through 2026.

The 2025 RealPage settlement required measures including restrictions on use of competitors' non-public information in runtime pricing, restrictions on certain training data, removal/redesign of certain pricing features, and monitoring.

Importance

This case demonstrates how investigators can examine an entire algorithmic ecosystem:

Competitors → Data → Algorithm → Pricing recommendation → Market outcome

The investigation therefore cannot focus only on the final price. It can also examine the information architecture behind the price.

15. Case 5 — Cornish-Adebiyi v. Caesars Entertainment

Jurisdiction: United States
Subject: Algorithmic hotel pricing

In 2024, the DOJ and FTC submitted a statement of interest concerning allegations involving hotel room pricing algorithms.

The agencies stated that businesses cannot use an algorithm to engage in conduct that would violate antitrust law if carried out directly by humans.

Importance

The case illustrates a fundamental principle for AI-assisted enforcement:

Software is not a legal shield.

If competitors exchange competitively sensitive information and use an algorithm to coordinate their pricing, investigators can examine the underlying human and technological arrangements.

16. Case 6 — Google Shopping

Case: Google and Alphabet v. European Commission
Case number: T-612/17
Subject: Algorithmic ranking and self-preferencing

The Google Shopping litigation concerned Google's treatment of its own comparison-shopping service within its general search results.

The European courts examined Google's search algorithms, ranking mechanisms and the effects of its conduct on competing comparison-shopping services. The case is therefore highly relevant to algorithmic antitrust investigations.

Importance

The case demonstrates how competition authorities can investigate:

• ranking algorithms;
• visibility of competitors;
• algorithmic demotion;
• preferential treatment;
• traffic diversion;
• effects on rival services.

This becomes increasingly relevant to AI systems that automatically rank, recommend, summarize, or distribute competing products.

17. Case 7 — U.S. v. Apple

Jurisdiction: United States
Filed: 2024
Subject: Digital ecosystem and monopolization

The DOJ's case against Apple concerns alleged monopolization and restraints involving Apple's ecosystem and digital services. The DOJ's case materials identify monopolization and other restraints among the asserted antitrust issues.

Although this is not an "AI investigation" case in the narrow sense, it is relevant to AI-assisted investigations because digital ecosystems can involve:

• APIs;
• interoperability;
• access restrictions;
• developer rules;
• data;
• platform distribution;
• technological dependencies.

These are also central issues in AI ecosystems.

18. Case 8 — Google Search / Digital Platform Investigations

The U.S. Google litigation provides another important example of how competition authorities investigate highly technical digital markets.

The investigation and litigation have involved large quantities of evidence concerning search distribution, agreements, data, technology and digital-market structures. The DOJ continues to maintain extensive case materials and enforcement proceedings concerning Google.

Relevance to AI

The investigative techniques developed in major digital-platform cases can be adapted to AI markets, particularly where competition depends upon:

• data;
• distribution;
• default settings;
• interoperability;
• cloud infrastructure;
• computing resources;
• platform access.

19. Role of Machine Learning in Investigations

A competition authority could theoretically train a system to identify patterns associated with previous cartel investigations.

For example:

Input

• millions of emails;
• millions of prices;
• contracts;
• company communications;
• transaction data.

AI processing

• clustering;
• anomaly detection;
• semantic analysis;
• network analysis;
• time-series analysis.

Output

• potentially relevant documents;
• suspicious transactions;
• unusual pricing patterns;
• possible communication networks.

Human stage

• investigators examine evidence;
• lawyers assess legal relevance;
• economists conduct economic analysis;
• authority determines whether formal proceedings are justified.

20. AI and Cartel Screening

A sophisticated cartel-screening system could search for:

Pricing indicators

• unusually parallel prices;
• simultaneous increases;
• unusual price stability.

Communication indicators

• frequent competitor communications;
• common intermediaries;
• unusual meeting patterns.

Data indicators

• sharing of future prices;
• sharing of capacity information;
• sharing of customer information.

Contract indicators

• market-allocation provisions;
• customer restrictions;
• bid coordination;
• resale-price provisions.

The AI output should normally be regarded as an investigative lead, not automatically as proof of infringement.

21. AI and Dawn Raids / Inspections

Competition authorities may obtain substantial electronic evidence during inspections.

AI can assist with:

• document classification;
• duplicate detection;
• entity recognition;
• communication mapping;
• identifying relevant keywords;
• prioritising documents;
• identifying relationships between employees and companies.

However, legal safeguards remain essential.

Authorities must respect applicable:

• procedural rights;
• confidentiality;
• legal professional privilege;
• privacy rules;
• search limitations;
• evidence requirements.

22. AI and Economic Analysis

AI can process very large datasets used by competition economists.

Possible applications include:

Demand estimation

AI can help analyse consumer responses to price changes.

Price elasticity

It can identify how demand changes when prices change.

Market-power analysis

It can assist in identifying whether a firm can profitably sustain prices above competitive levels.

Entry analysis

It can analyse historical entry and exit patterns.

Merger analysis

It can identify overlaps between products and customers.

23. AI and Merger Investigations

AI can assist authorities in reviewing proposed mergers involving AI companies.

Important questions include:

• Will the merger combine important datasets?
• Will it remove an emerging competitor?
• Will it strengthen control over computing infrastructure?
• Will the merged firm control an important distribution channel?
• Will competitors lose access to technology?
• Will the transaction increase switching costs?
• Will it increase vertical foreclosure risks?

The FTC, for example, used compulsory Section 6(b) orders in 2024 to examine generative-AI partnerships and investments involving major AI and cloud companies.

24. AI Infrastructure and Competition

AI competition is not limited to applications.

Investigators may examine:

Chips → Computing → Cloud → Foundation models → Applications → Distribution

Potential competition concerns can arise if one company controls an important bottleneck.

The FTC has specifically discussed potential bottlenecks involving cloud infrastructure and access to GPUs in the AI ecosystem.

The OECD likewise identifies AI infrastructure as an area requiring competition scrutiny, while noting that completed enforcement specifically concerning AI infrastructure remains limited.

25. AI-Assisted Investigation Under EU Competition Law

The principal provisions include:

Article 101 TFEU

Relevant where AI-enabled arrangements may involve:

• price fixing;
• market sharing;
• output restrictions;
• exchange of competitively sensitive information;
• other restrictive agreements or concerted practices.

Article 102 TFEU

Relevant where a dominant undertaking uses AI or technological control to engage in exclusionary or exploitative conduct.

The Commission can initiate Article 102 investigations following complaints, on its own initiative, or through sector inquiries.

26. AI-Assisted Investigation Under U.S. Antitrust Law

The main statutes include:

Sherman Act §1

Relevant to agreements and coordinated conduct.

Sherman Act §2

Relevant to monopolization and attempted monopolization.

FTC Act §5

Provides an additional enforcement framework for the FTC.

The important point is that the technology used to implement conduct does not fundamentally change the underlying competition-law analysis.

27. AI-Assisted Investigation Under Indian Competition Law

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

Important provisions include:

Section 3

Deals with anti-competitive agreements.

Section 4

Deals with abuse of dominant position.

Section 5

Deals with combinations.

Section 19

Provides the framework for inquiry into certain agreements and dominant-position cases.

Section 26

Deals with the investigation process following formation of a prima facie opinion.

AI can assist the CCI in identifying potential issues, but the statutory investigation and adjudicatory process remains governed by the Competition Act and applicable procedural safeguards.

The Samir Agrawal/Ola-Uber litigation is particularly relevant because it directly considered allegations concerning algorithmic pricing.

28. Evidentiary Problems With AI-Assisted Investigations

AI-assisted investigations create several legal challenges.

A. Explainability

If an AI system identifies a company as suspicious, investigators should be able to understand why.

B. False positives

A legitimate competitive strategy may resemble cartel behaviour.

C. False negatives

AI may fail to detect sophisticated coordination.

D. Bias

Training data may contain historical biases.

E. Data quality

Poor data can produce unreliable conclusions.

F. Confidentiality

Competition investigations often contain commercially sensitive information.

G. Due process

Businesses should have meaningful opportunities to challenge evidence used against them.

29. Human Oversight

A sound investigative framework should therefore look like:

AI detection

↓

Human verification

↓

Economic analysis

↓

Legal analysis

↓

Collection of supporting evidence

↓

Procedural safeguards

↓

Final enforcement decision

The AI system should not simply determine that:

"Company X violated competition law."

That conclusion requires legal and evidentiary assessment by the competent authority or court.

30. Key Legal Questions for Future AI Antitrust Cases

Future cases are likely to examine questions such as:

1. When does algorithmic pricing become unlawful coordination?
2. Can autonomous AI agents independently reach an unlawful coordinated outcome?
3. Who is legally responsible for an AI agent's conduct?
4. Can a common AI provider become a hub for competitor coordination?
5. When does sharing training data become an antitrust concern?
6. Can dominant AI companies deny rivals access to essential inputs?
7. Can AI ranking systems unlawfully favour a platform's own products?
8. Can AI systems facilitate discriminatory exclusion?
9. How should authorities prove causation between an algorithm and market harm?
10. What level of transparency should authorities require from AI systems?

31. Important Distinction: AI Investigation vs. AI Antitrust Violation

These concepts should not be confused.

The same technology can therefore be both:

a tool for enforcement and the subject of enforcement.

32. Major Challenges for Competition Authorities

1. Technical complexity

Investigators need technical knowledge of:

• machine learning;
• model architecture;
• APIs;
• data pipelines;
• cloud computing;
• algorithmic pricing.

2. Rapid technological development

AI systems can change faster than traditional regulatory investigations.

3. Lack of transparency

Some AI systems are difficult to explain.

4. Cross-border conduct

AI platforms operate internationally.

5. Massive datasets

Traditional investigative methods may be insufficient for very large datasets.

6. Autonomous systems

Future AI agents may make pricing, purchasing, advertising, or distribution decisions with limited human intervention.

33. Compliance Implications for Businesses

Businesses using AI should consider:

• documenting algorithm design;
• maintaining competition-law policies;
• controlling access to competitor data;
• auditing pricing algorithms;
• monitoring third-party AI providers;
• documenting independent pricing decisions;
• avoiding inappropriate competitor information exchanges;
• reviewing common algorithms used by competitors;
• maintaining records of model changes;
• establishing human oversight.

The purpose is not to prevent businesses from using AI, but to ensure that technological systems do not inadvertently facilitate prohibited conduct.

34. Overall Legal Framework

The developing framework can be summarised as:

Traditional competition law

Digital-market enforcement

Algorithmic evidence

AI/data analytics

Human legal and economic assessment

=

AI-assisted antitrust enforcement

Recent international enforcement activity shows that authorities are increasingly focusing on AI-related competition risks. In July 2024, the FTC, DOJ and international competition authorities issued a joint statement addressing competition risks in AI markets.

The European Commission has also identified AI and cloud services as important areas for future digital-market scrutiny.

35. Conclusion

AI-assisted antitrust investigations represent an evolution in enforcement methodology rather than a completely new branch of competition law.

AI can help competition authorities process enormous datasets, identify suspicious conduct, analyse pricing patterns, examine communications, understand digital ecosystems and investigate potentially anti-competitive behaviour.

The cases of Topkins, Eturas, Samir Agrawal, RealPage, Cornish-Adebiyi and Google Shopping demonstrate different aspects of the legal problem: algorithmic price coordination, automated restrictions, evidentiary requirements, algorithmic information sharing, and algorithmic ranking.

The central principle emerging from these authorities is that competition law focuses on the substance and effects of business conduct, not merely on whether a human or an algorithm performed the relevant action. At the same time, an AI-generated pattern or prediction is not by itself proof of an infringement; authorities must establish the applicable legal elements using reliable evidence and respect procedural safeguards.