Competition Law and AI-Assisted Antitrust Investigations

1. Introduction

Artificial Intelligence (AI) is increasingly being used by competition authorities to assist with antitrust investigations, particularly where authorities must analyse very large volumes of emails, contracts, transaction records, pricing data, algorithms, platform data, and communications.

AI does not replace the legal decision-maker. Its main role is investigative and analytical: identifying patterns, prioritising evidence, detecting potentially coordinated conduct, mapping business relationships, and helping investigators formulate questions that can then be tested using conventional evidence.

The use of AI is especially relevant in digital markets, where competitive conduct may depend on algorithms, automated pricing, recommender systems, ranking systems, APIs, data access, cloud infrastructure, and AI models.

The CMA, for example, has specifically examined competition risks in AI foundation-model markets, including risks arising from access to data, computing infrastructure, partnerships and distribution channels. The FTC, DOJ and international competition authorities have likewise identified competition concerns across the AI ecosystem.

2. Meaning of AI-Assisted Antitrust Investigation

An AI-assisted antitrust investigation is an investigation in which competition authorities use machine-learning, natural-language processing, data analytics or other AI techniques to assist in discovering, organising or analysing evidence relevant to competition law.

It can be used for:

1. Document review
2. Email and communication analysis
3. Contract analysis
4. Price-pattern analysis
5. Bid-rigging detection
6. Market-definition analysis
7. Network analysis
8. Algorithmic conduct investigation
9. Merger screening
10. Detection of exclusionary conduct
11. Identification of potentially coordinated behaviour
12. Analysis of large digital-platform datasets

The final legal conclusion, however, must remain based on applicable competition law and admissible/reliable evidence rather than an unexplained AI output.

3. Why AI Is Important in Antitrust Investigations

Traditional investigations can involve millions of documents and enormous datasets.

For example, an investigation may contain:

• millions of emails;
• instant messages;
• internal documents;
• pricing records;
• customer information;
• bidding data;
• contracts;
• source-code records;
• algorithmic logs;
• platform-ranking information;
• transaction data;
• cloud-computing records.

Human investigators cannot manually examine every piece of information.

AI can therefore act as an investigative filtering and analytical tool.

Example

Suppose ten companies participate in an online procurement market.

An AI system could analyse:

• millions of historical prices;
• bidding patterns;
• timing of bids;
• communications;
• geographic allocation;
• customer allocation;
• repeated pricing patterns.

The AI might identify an unusual pattern.

That pattern would not itself prove cartel conduct. Investigators would then examine underlying documents and other evidence to determine whether the pattern has a legitimate explanation or supports a competition-law theory.

4. Main Areas of AI-Assisted Antitrust Investigation

A. Document Discovery

AI can classify large collections of documents according to relevance.

For example, an authority investigating a suspected cartel could use natural-language processing to identify documents relating to:

• price discussions;
• customer allocation;
• tender strategy;
• competitors;
• discounts;
• market sharing;
• production restrictions.

This can significantly reduce the amount of material investigators need to examine manually.

B. Email and Communication Analysis

AI can identify relationships among individuals and organisations.

For example, network analysis may reveal:

Company A → Distributor X → Company B → Competitor C

Investigators can then determine whether the relationships have legitimate commercial explanations or whether they are relevant to suspected coordination.

AI can also identify:

• unusual communication frequency;
• common terminology;
• sudden communication between competitors;
• clusters of employees;
• repeated references to particular customers;
• communications occurring shortly before pricing changes.

Again, these are investigative indicators rather than automatic findings of liability.

5. AI and Cartel Detection

Cartels are traditionally investigated through evidence such as:

• communications;
• meetings;
• agreements;
• pricing patterns;
• bid allocation;
• witness testimony;
• leniency applications.

AI can add statistical and computational analysis.

Potential indicators

AI systems may identify:

• parallel price movements;
• identical or unusually similar bids;
• suspicious bid rotation;
• geographic allocation patterns;
• unusually stable market shares;
• coordinated supply restrictions;
• repeated timing patterns.

However, parallel behaviour alone is not necessarily unlawful coordination.

Competition law normally requires the authority to establish the legally relevant elements of the infringement.

6. Algorithmic Pricing and Tacit Coordination

One of the most important issues is algorithmic pricing.

Companies increasingly use software to determine prices automatically.

This creates a difficult question:

When several companies use algorithms that respond to competitors' prices, can the resulting coordination amount to an antitrust infringement?

The answer depends on the applicable jurisdiction and facts.

An authority may investigate:

• the algorithm's design;
• instructions given to developers;
• pricing objectives;
• information available to the algorithm;
• communication between competitors;
• whether firms intentionally configured systems to coordinate;
• whether the resulting conduct is independent or the result of an agreement or concerted practice.

AI can help authorities reconstruct historical pricing behaviour and identify whether changes occurred systematically.

7. AI and Market Definition

Market definition is fundamental to many competition-law investigations.

AI can analyse:

• consumer-substitution data;
• search behaviour;
• purchasing patterns;
• product characteristics;
• geographic transactions;
• switching behaviour;
• platform usage.

For example, investigators could examine whether consumers regard two AI services as substitutes.

Possible questions include:

• Are general search engines substitutes for AI assistants?
• Are cloud AI services substitutes across providers?
• Are open-source and proprietary models competing products?
• Is AI infrastructure a separate market?
• Does access to specialised computing constitute a separate competitive constraint?

These questions must ultimately be answered using established competition-law methodology rather than an AI-generated conclusion.

8. AI and Abuse of Dominance

AI can assist investigations involving potentially abusive conduct such as:

Self-preferencing

A dominant platform may allegedly favour its own AI service over competing AI services.

Tying

A company may connect an AI service to another product or platform in a manner that raises competition concerns.

Exclusive dealing

AI developers may be restricted from using competing infrastructure or distribution channels.

Refusal to deal

Access to essential data, APIs, computing resources or infrastructure may become an investigative issue.

Discriminatory access

A platform may provide different technical access to its own AI service and competitors.

The European Commission's 2026 DMA work concerning Google illustrates how AI-related interoperability and access to search data have become concrete regulatory issues. The Commission adopted measures concerning interoperability for competing AI services on Android and access to Google Search data.

9. AI and Merger Investigations

AI can also assist merger investigations.

Competition authorities may use AI to examine:

• overlapping products;
• customer relationships;
• competitors;
• patent portfolios;
• datasets;
• cloud infrastructure;
• AI models;
• employees and specialised talent;
• distribution networks.

This is particularly relevant where a transaction involves:

AI model + cloud infrastructure + data + distribution platform.

The authority may need to determine whether the transaction could eliminate an emerging competitor or strengthen an existing ecosystem.

10. AI and Evidence

This is one of the most important legal issues.

AI-generated analysis should generally be treated as an investigative aid, not automatically as proof.

A competition authority should be able to explain:

1. what data was used;
2. how the data was collected;
3. how the AI system processed it;
4. what assumptions were used;
5. whether the dataset was complete;
6. whether the model produced false positives;
7. whether investigators independently verified the output.

This is particularly important because AI systems can produce:

• false positives;
• false negatives;
• biased results;
• incomplete classifications;
• hallucinated information;
• unexplained correlations.

11. Six Important Case Laws

There is currently a limited body of reported judicial decisions specifically concerning AI-assisted antitrust investigations. Therefore, the most useful case law consists of established competition cases dealing with algorithmic conduct, digital evidence, information exchange, pricing coordination and digital-platform behaviour.

Case 1 — Eturas UAB v Lietuvos Respublikos Konkurencijos Taryba

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

Facts

Eturas operated an online travel-booking system. The system included a mechanism that could restrict discounts offered by travel agencies using the platform.

The case concerned whether participating businesses could be held responsible for anticompetitive coordination where information was communicated through the platform.

Legal importance

The CJEU considered the evidentiary significance of an electronic message distributed through a common platform.

The case is particularly relevant to AI-assisted investigations because it demonstrates that competition authorities may need to examine:

• digital communications;
• platform architecture;
• automated mechanisms;
• knowledge of participants;
• conduct following electronic communications.

Relevance to AI

AI could help investigators identify similar communications and determine which businesses received them and what conduct followed.

Principle: Digital evidence can be important in establishing or disproving concerted practices, but the legal inference must be supported by the surrounding evidence.

Case 2 — United States v Apple Inc.

Court: United States District Court, Southern District of New York
Year: 2024

Facts

The U.S. Department of Justice brought an antitrust action concerning Apple's conduct in smartphone markets.

The case involved allegations concerning Apple's control over its ecosystem and the effect of its practices on competition.

Relevance to AI-assisted investigations

Large technology investigations require examination of:

• internal communications;
• product strategies;
• technical documents;
• APIs;
• contracts;
• developer relationships;
• platform restrictions.

AI tools can help investigators organise and analyse such large evidence collections.

Principle

Digital-platform antitrust investigations require analysis of both commercial conduct and technical architecture.

This becomes even more important when platforms incorporate AI services into operating systems and ecosystems.

Case 3 — Google Search (Shopping)

Case: Google Search (Shopping)
European Commission Decision: 2017
General Court: Google and Alphabet v Commission, Joined Cases T-612/17 and related proceedings

Facts

The European Commission found that Google had given prominent placement to its comparison-shopping service while applying less favourable treatment to competing comparison-shopping services.

The General Court largely upheld the Commission's findings, subject to certain modifications concerning aspects of the analysis.

Relevance to AI

The case is highly relevant to modern AI-platform investigations because ranking systems can influence which competitors consumers see.

AI-assisted investigations may examine:

• ranking data;
• search results;
• recommendation systems;
• visibility of competitors;
• changes in ranking algorithms;
• treatment of proprietary services.

The Commission continues to address self-preferencing and platform-access issues under the DMA. In 2026, it imposed fines on Google concerning self-preferencing in Search and anti-steering restrictions in Google Play.

Principle

Algorithmic ranking is not outside competition law simply because the ranking process is technically complex.

Case 4 — Intel Corp. v Commission

Court: Court of Justice of the European Union
Case: C-413/14 P
Judgment: 2017

Facts

The case concerned rebates granted by Intel and the assessment of whether those rebates could constitute abusive exclusionary conduct.

Importance for AI-assisted investigations

The case is important for understanding how competition authorities must analyse economic evidence rather than relying solely on the existence of a particular commercial practice.

AI could assist with:

• customer-level data;
• pricing records;
• rebate calculations;
• competitor coverage;
• economic datasets.

But the legal analysis must still establish the relevant competitive effects using appropriate methodology.

Principle

Complex economic evidence requires careful analysis of the actual competitive circumstances rather than an automatic assumption based on the form of conduct.

Case 5 — Qualcomm v Commission

Court: General Court of the European Union
Case: T-235/18
Judgment: 2022

Facts

The European Commission had found that Qualcomm had made payments to Apple under arrangements concerning the supply of LTE chipsets.

The General Court annulled the Commission's decision, identifying shortcomings in the Commission's assessment, including aspects of its analysis of exclusionary effects and procedure.

Relevance to AI

This case demonstrates an important limitation on AI-assisted investigation:

More data does not automatically mean better legal analysis.

An AI system may identify a pattern, but authorities must still:

• establish the correct legal test;
• analyse relevant evidence;
• consider countervailing evidence;
• respect procedural requirements;
• provide adequate reasoning.

Principle

Competition investigations require legally and economically rigorous reasoning, not merely large-scale data analysis.

Case 6 — Google Android

Case: Google Android
European Commission Decision: 2018
General Court: Case T-604/18, Google and Alphabet v Commission
Judgment: 2022

Facts

The Commission investigated Google's Android practices involving:

• search;
• app stores;
• mobile-device manufacturers;
• licensing arrangements.

The Commission found several practices to be abusive, and the General Court substantially upheld the Commission's findings while reducing the fine.

Relevance to AI

AI services are increasingly integrated into operating systems and app ecosystems.

Therefore, the Android litigation provides a useful legal framework for examining future issues involving:

• AI assistants;
• operating-system access;
• default settings;
• app distribution;
• interoperability;
• competing AI applications.

In 2026, the European Commission specifically adopted DMA measures addressing interoperability between competing AI services and Android functionality.

12. Additional Relevant Case — T-Mobile Netherlands

Case: C-8/08
Court: Court of Justice of the European Union
Judgment: 2009

The case concerned information exchange between competitors.

The CJEU examined when exchanges of information can reduce strategic uncertainty and constitute a restriction of competition.

AI relevance

AI can make information exchange more powerful because algorithms can:

• collect market information;
• process competitors' prices;
• update prices rapidly;
• respond automatically to market changes.

Therefore, investigators must distinguish between:

Independent algorithmic behaviour

and

algorithmic implementation of an anticompetitive agreement or coordinated practice.

13. AI as an Investigative Tool

A competition authority can potentially use AI at several stages.

14. AI and Dawn Raids

Competition authorities sometimes conduct inspections of business premises and digital systems.

Modern inspections can involve:

• computers;
• cloud accounts;
• mobile devices;
• email servers;
• collaboration platforms;
• databases;
• source code;
• algorithmic logs.

AI can assist investigators in sorting and prioritising the enormous amount of information collected.

However, investigators must respect applicable rules concerning:

• legal privilege;
• privacy;
• confidentiality;
• relevance;
• proportionality;
• procedural fairness;
• data protection.

15. AI and Leniency Investigations

Leniency programmes allow cartel participants to cooperate with competition authorities.

AI could assist authorities in comparing information supplied by different applicants.

For example:

Company A statement

↓

AI identifies alleged meeting dates

↓

Company B documents

↓

AI identifies matching communications

↓

Company C records

↓

AI identifies matching pricing events

The investigators would then manually verify the evidence.

AI therefore functions as an evidence-linking mechanism, not as the final adjudicator.

16. AI and Economic Evidence

AI can process large economic datasets.

For example, investigators could analyse:

• price elasticity;
• margins;
• market shares;
• customer switching;
• geographic sales;
• transaction frequency;
• discount patterns.

Machine-learning techniques may identify relationships that conventional analysis might initially overlook.

However, correlation does not necessarily establish causation.

This distinction is critical in competition law.

17. Problems and Risks

A. Black-box decision-making

If investigators cannot understand why an AI system identified a particular company or transaction as suspicious, its evidentiary value may be limited.

B. False positives

An algorithm may classify legitimate business behaviour as suspicious.

C. False negatives

The system may fail to detect sophisticated coordination.

D. Data bias

Incomplete or biased datasets can distort the investigation.

E. Privacy

Large-scale analysis may involve personal information.

F. Legal privilege

AI systems must not improperly expose privileged communications.

G. Procedural fairness

Businesses must have an opportunity to understand and challenge the evidence relied upon against them.

H. Automation bias

Investigators may give excessive weight to an AI-generated result simply because it appears technically sophisticated.

18. Human Oversight

A sound AI-assisted investigation should follow a structure such as:

AI identification → Human verification → Legal analysis → Economic analysis → Procedural review → Final decision

Rather than:

AI prediction → Automatic finding of infringement

The second approach creates serious risks for due process.

19. AI-Specific Competition Concerns

AI itself can create new competition problems.

1. Compute concentration

A small number of firms may control access to advanced computing resources.

2. Data concentration

Large datasets may provide significant competitive advantages.

3. Foundation-model concentration

A small number of companies may develop commercially important foundation models.

4. Cloud dependency

AI developers may depend heavily on a limited number of cloud providers.

5. Exclusive partnerships

Strategic partnerships between AI developers and major technology platforms may affect market access.

6. Distribution control

Operating systems and app stores can influence the distribution of AI assistants.

7. Self-preferencing

A platform may give preferential treatment to its own AI product.

8. Interoperability restrictions

Technical restrictions may make it harder for competing AI services to access platform functionality.

The European Commission's 2026 Android interoperability measures specifically address this type of competition concern, seeking effective access for competing AI services to Android capabilities.

20. AI Foundation Models and Competition Law

The UK CMA's AI Foundation Models review identified competition risks associated with the development and deployment of foundation models and subsequently established principles intended to support competitive AI markets.

Important areas include:

• access to critical inputs;
• access to computing power;
• partnerships;
• distribution;
• technical interoperability;
• switching;
• data advantages;
• vertical integration;
• market concentration.

These issues can potentially arise under:

• abuse-of-dominance rules;
• merger control;
• agreements/concerted-practice rules;
• sector-specific digital regulation.

21. Relationship Between AI and Traditional Competition Law

AI does not create a completely separate competition-law universe.

Existing doctrines remain relevant.

Article 101 TFEU / cartel rules

Relevant where AI facilitates:

• price fixing;
• market sharing;
• bid rigging;
• information exchange;
• coordination.

Article 102 TFEU / abuse of dominance

Relevant where dominant firms use AI-related assets to:

• exclude competitors;
• discriminate;
• self-preference;
• tie products;
• restrict access.

Merger control

Relevant where AI-related transactions may substantially lessen or eliminate competition.

Digital regulation

The EU DMA increasingly provides additional obligations for designated gatekeepers. The Commission has also been using DMA procedures to address AI-related interoperability and data-access questions.

22. Investigative Framework

A competition authority using AI should ideally follow these stages:

Stage 1 — Define the legal question

Determine whether the investigation concerns:

• cartel;
• abuse of dominance;
• merger;
• vertical restriction;
• information exchange;
• digital-platform conduct.

Stage 2 — Identify relevant datasets

Collect:

• documents;
• emails;
• transaction records;
• prices;
• contracts;
• algorithmic logs;
• platform data.

Stage 3 — AI-assisted screening

Use AI to identify:

• relevant documents;
• unusual patterns;
• relationships;
• anomalies.

Stage 4 — Human verification

Investigators independently examine the relevant evidence.

Stage 5 — Economic analysis

Economists assess:

• market definition;
• market power;
• effects;
• counterfactuals;
• efficiencies.

Stage 6 — Legal analysis

Lawyers apply the applicable competition-law test.

Stage 7 — Procedural safeguards

The authority considers:

• confidentiality;
• privilege;
• disclosure;
• rights of defence;
• reliability of evidence.

Stage 8 — Final decision

The final decision must be based on legally sufficient evidence and reasoning, rather than simply on an AI-generated score.

23. Important Legal Principle

The most important principle is:

AI can assist an antitrust investigation, but AI should not become a substitute for legal proof.

An algorithm can tell investigators:

"This transaction pattern is unusual."

It cannot automatically establish:

"The company violated competition law."

The latter requires application of the relevant legal standard to verified evidence.

24. Future Development

AI-assisted antitrust investigations are likely to become increasingly important because competition authorities are dealing with markets in which:

• algorithms determine prices;
• AI systems determine rankings;
• platforms control distribution;
• data provides competitive advantages;
• cloud infrastructure supports AI models;
• automated systems interact with one another.

The FTC, DOJ, CMA and European Commission have publicly recognised that AI can create significant competition issues and have emphasised continued enforcement and cooperation in the AI ecosystem.

The European Commission's recent work also shows that AI competition questions are moving from theoretical discussion toward concrete regulatory issues, particularly around AI interoperability, search data and platform access.

25. Conclusion

Competition Law and AI-Assisted Antitrust Investigations represents the intersection of traditional antitrust principles and modern computational investigation.

AI can significantly improve the ability of competition authorities to:

• process massive datasets;
• detect suspicious patterns;
• analyse digital communications;
• investigate algorithms;
• examine pricing behaviour;
• identify market relationships;
• analyse platform ecosystems;
• support merger investigations.

However, AI-generated findings must be treated carefully. The cases involving Eturas, Intel, Qualcomm, Google Shopping, Google Android and T-Mobile Netherlands demonstrate important principles concerning digital evidence, economic analysis, information exchange, platform conduct and procedural safeguards.

The emerging legal model is therefore best understood as:

AI-assisted discovery + human verification + economic analysis + legal reasoning + procedural safeguards.

That approach allows competition authorities to use AI's analytical capabilities while preserving the fundamental requirements of evidence, due process and reasoned antitrust decision-making.