Competition Law and AI-Controlled Corporate Groups

Competition Law and AI-Controlled Corporate Groups

1. Introduction

An AI-controlled corporate group can be understood as a group of legally separate companies in which artificial-intelligence systems play an important role in coordinating or influencing decisions across the parent company, subsidiaries, affiliates, or business divisions.

For example, a corporate group might use a central AI system to determine:

• prices charged by several subsidiaries;
• purchasing and procurement policies;
• allocation of customers or territories;
• production levels;
• advertising and distribution strategies;
• access to data, cloud infrastructure, or digital platforms;
• decisions about suppliers and competitors;
• internal investment priorities; and
• commercial terms offered by different companies in the group.

Competition law does not normally treat AI itself as an independent legal person or undertaking. Instead, authorities generally examine the companies and economic units that design, deploy, control, benefit from, or implement the AI-assisted conduct.

In the European Union, the starting points remain Articles 101 and 102 TFEU. Article 101 prohibits certain anticompetitive agreements and concerted practices, while Article 102 prohibits abuse by undertakings holding a dominant position.

There is not yet a large body of decided cases specifically concerning an autonomous AI system controlling an entire corporate group. Therefore, the most relevant legal principles come from established cases involving parent-subsidiary liability, single economic units, digital ecosystems, algorithms, dominance and group responsibility.

2. The Concept of an “Undertaking”

Competition law focuses primarily on the undertaking, rather than simply on the individual corporation.

An undertaking is essentially an economic entity engaged in economic activity. Several legally separate companies can therefore constitute one undertaking when they operate as a single economic unit.

This principle becomes particularly important for AI-controlled corporate groups.

Suppose:

Parent Company P owns Subsidiaries A, B and C.

A central AI platform determines important commercial decisions for all four companies.

The legal question is not simply:

Which legal company operated the AI?

The authority may instead investigate whether P, A, B and C constitute one economic unit for the activity concerned.

EU case law confirms that formal separation between companies does not necessarily prevent them from constituting one undertaking for competition-law purposes.

3. AI Does Not Automatically Remove Corporate Responsibility

A company would generally have difficulty avoiding competition-law responsibility merely by arguing:

“The algorithm made the decision.”

Competition law examines the economic conduct of the undertaking.

Therefore, relevant questions could include:

• Who selected the AI system?
• Who established its commercial objectives?
• What data was supplied to it?
• Which companies followed its recommendations?
• Could management override its decisions?
• Was the AI centrally operated by the parent company?
• Were subsidiaries commercially independent?
• Did the AI coordinate prices or output?
• Did management know how the system affected competition?

The degree of human involvement could matter to particular questions of evidence or liability, but automation does not automatically place commercial conduct outside competition law.

4. Single Economic Unit Doctrine

The single economic unit doctrine is particularly important.

A parent company and subsidiary can constitute one undertaking where the subsidiary does not independently determine its market conduct but essentially follows the direction or decisive influence of its parent.

AI could make this relationship more complicated.

Imagine that a parent company operates a central AI system controlling:

• subsidiary pricing;
• inventory;
• supplier selection;
• customer allocation;
• advertising;
• investment; and
• market-entry decisions.

Even without daily instructions from human executives, centralized algorithmic control could potentially provide evidence relevant to determining whether subsidiaries possess genuine commercial autonomy.

The precise legal conclusion would depend on the facts.

5. Parent Company Liability

EU competition law permits responsibility in appropriate circumstances to extend beyond the subsidiary that directly carried out the infringement.

Where a parent exercises decisive influence over its subsidiary, they may form part of the same undertaking.

This matters considerably for AI-controlled groups.

Suppose Subsidiary A operates an AI pricing system that engages in anticompetitive conduct.

Investigators may examine whether:

1. the parent controlled A;
2. the parent controlled the AI infrastructure;
3. the parent established the AI's objectives;
4. group-wide data was supplied to the system;
5. the parent supervised implementation; and
6. A had genuine independence over its commercial policy.

Thus, liability cannot necessarily be isolated inside whichever subsidiary technically owns the relevant software.

6. Wholly Owned Subsidiaries

An especially important rule arises where a parent owns all or virtually all of a subsidiary.

EU case law recognizes a rebuttable presumption, in the relevant circumstances, that a parent holding all or virtually all of the subsidiary's capital exercises decisive influence over that subsidiary.

The Court has continued to describe this principle in recent case law.

Therefore, placing an AI system in a wholly owned technology subsidiary would not necessarily insulate the remainder of the corporate group from competition-law consequences.

7. AI-Controlled Pricing

One major risk concerns algorithmic pricing.

A corporate group could deploy AI capable of continuously adjusting prices according to:

• competitor prices;
• demand;
• consumer behaviour;
• inventory;
• geographical location;
• market conditions; and
• historical transactions.

Using algorithms to optimize prices is not inherently unlawful.

The competition concern arises where the surrounding arrangement results in conduct such as:

• price fixing between independent competitors;
• coordination of commercially sensitive information;
• market allocation;
• output restrictions; or
• other anticompetitive coordination.

Article 101 expressly covers practices including price fixing, limiting production or markets, and market sharing.

8. Intra-Group Coordination Versus Coordination Between Competitors

An important distinction must be made.

If Subsidiaries A and B belong to the same undertaking, coordination between them generally is not treated in the same way as an agreement between independent competing undertakings under Article 101.

But suppose the group's AI system communicates with or receives competitively sensitive information from independent competitors.

The situation becomes substantially different.

For example:

Group X AI → Competitor Y AI → Competitor Z AI

If those systems facilitate coordination of prices, customers or output between independent undertakings, Article 101 concerns could arise.

The legal analysis therefore requires identifying the boundaries of the relevant undertaking.

9. AI and Abuse of Dominance

AI-controlled corporate groups can also create issues under Article 102 TFEU.

Article 102 applies where an undertaking holds a dominant position and abuses that position.

Dominance itself is not prohibited.

The concern is abusive conduct.

The Commission's Article 102 framework addresses exclusionary behaviour by dominant undertakings, and in September 2026 the Commission adopted its first Guidelines concerning abusive exclusionary conduct.

AI could potentially be involved in conduct concerning:

• discriminatory access;
• tying;
• exclusivity;
• self-preferencing;
• refusal of access;
• interoperability restrictions;
• exclusionary pricing;
• platform access conditions; or
• restrictions affecting technological development.

Whether particular conduct violates Article 102 depends on the legal and economic circumstances.

10. AI-Controlled Digital Ecosystems

Modern corporate groups increasingly operate interconnected ecosystems involving:

AI + cloud computing + data + platforms + applications + advertising + payment systems + devices.

Control over several complementary markets can create complicated competition issues.

A central AI system could decide which group products receive preferential:

• ranking;
• recommendation;
• advertising;
• interoperability;
• data access; or
• platform placement.

Competition authorities would therefore examine the economic effects rather than merely the software architecture.

The Google Android litigation demonstrates how competition analysis can cover an integrated ecosystem involving operating systems, app stores, search applications, browsers and contractual restrictions.

11. Data Concentration

AI-controlled groups can depend heavily on large datasets.

Suppose a corporate group controls:

• consumer data;
• transaction data;
• advertising information;
• logistics information;
• search information; and
• AI training infrastructure.

A competition investigation might examine whether control over those assets contributes to market power or whether particular conduct concerning those assets excludes competitors.

Possession of extensive data is not itself automatically unlawful.

The analysis concerns matters such as market structure, substitutability, barriers to entry and the conduct of the undertaking.

12. AI and Corporate Acquisitions

AI also matters in merger control.

Large corporate groups may acquire:

• AI startups;
• model developers;
• data companies;
• cloud infrastructure businesses;
• semiconductor companies; or
• specialized software suppliers.

Authorities can examine whether an acquisition may significantly reduce competition.

This is increasingly important where the target is currently small but possesses strategically important technology, data, talent or innovation capacity.

The European Commission published draft revised Merger Guidelines on 30 April 2026, reflecting major changes including digitalisation and new market realities.

13. AI and Information Exchange

Another risk concerns information sharing.

Imagine several independent companies use the same AI service.

Each company supplies:

• future prices;
• inventory information;
• sales forecasts;
• production plans; and
• customer information.

If commercially sensitive information is improperly pooled or used to coordinate competitors' behaviour, competition concerns may arise.

The important issue is not simply that a common algorithm exists.

Investigators would examine the nature of the information exchange and whether it reduces competitive uncertainty between independent undertakings.

14. Compliance Implications

Corporate groups using centralized AI should therefore maintain clear governance over important competition-sensitive functions.

Useful controls can include:

• identifying which companies use particular AI systems;
• documenting who determines commercial objectives;
• controlling access to competitors' confidential information;
• monitoring pricing systems;
• maintaining appropriate human oversight;
• auditing significant algorithmic decisions;
• establishing competition-law compliance rules for AI systems; and
• preserving appropriate records explaining important commercial decisions.

An AI system should not become an unmonitored mechanism through which potentially unlawful coordination or exclusionary conduct occurs.

Important Case Laws

Case 1 — Akzo Nobel and Others v Commission

Case C-97/08 P, Court of Justice, 2009

This is one of the foundational EU cases concerning parent-company responsibility.

The Court confirmed the importance of the concept of decisive influence.

Where a parent owns 100% of a subsidiary that has infringed EU competition rules, there is a rebuttable presumption that the parent exercises decisive influence over the subsidiary.

Importance for AI-Controlled Groups

Suppose an AI subsidiary is wholly owned by a technology holding company.

The holding company cannot necessarily escape responsibility simply because the AI software and employees are legally located in the subsidiary.

If the entities constitute one undertaking, competition liability can extend through the corporate structure.

The decisive-influence principle remains expressly recognized in subsequent Court case law.

Case 2 — Imperial Chemical Industries v Commission

Case 48/69, Court of Justice, 1972

This early case established an important principle concerning corporate groups.

Separate legal personality does not necessarily mean separate economic behaviour for competition-law purposes.

A subsidiary may lack genuine autonomy where the parent exercises sufficient influence over its market conduct.

Importance for AI

Imagine subsidiaries are formally managed by separate boards but their commercial strategies are effectively determined by the same centralized AI infrastructure.

A competition authority could examine the actual economic relationship rather than stopping at formal corporate boundaries.

Recent Court materials continue to cite ICI for the proposition that formal legal separation does not necessarily preclude unity of market conduct.

Case 3 — Skanska Industrial Solutions and Others

Case C-724/17, Court of Justice, 2019

Skanska is particularly significant because the Court emphasized the EU competition-law concept of the undertaking in the context of liability for competition infringements.

Corporate restructuring cannot necessarily eliminate responsibility connected with the relevant economic entity.

Importance for AI

Suppose a corporate group:

1. operates an AI system involved in anticompetitive conduct;
2. transfers that AI business to another group subsidiary;
3. dissolves the original company; and
4. continues essentially the same economic activity.

Competition-law responsibility cannot necessarily be avoided simply through changes in corporate legal structure.

The Court continues to rely on Skanska when explaining the functional nature of the undertaking concept.

Case 4 — Sumal SL v Mercedes Benz Trucks España SL

Case C-882/19, Court of Justice, 2021

Sumal significantly developed the single-economic-unit doctrine.

The Court explained that liability can, under appropriate conditions, operate in relation to a subsidiary where the parent and subsidiary belong to the same economic unit and there is a concrete link between the subsidiary's economic activity and the subject matter of the infringement.

The Court also explained that the relevant economic unit is connected to the activity concerned rather than simply to the corporate group's overall legal structure.

Importance for AI Groups

This is particularly relevant where a multinational group has numerous subsidiaries using a common AI platform.

Membership in the same corporate group alone does not mean every subsidiary automatically becomes liable for every infringement.

The economic activity and relationship to the infringement remain important.

Case 5 — Google and Alphabet v Commission (Google Android)

Case T-604/18, General Court, 2022; subsequent appeal C-738/22 P

This case concerned Google's Android ecosystem.

The litigation involved:

• Android;
• the Play Store;
• Google Search;
• Chrome;
• device manufacturers;
• mobile network operators;
• product bundles;
• exclusivity arrangements; and
• anti-fragmentation obligations.

The General Court largely upheld the Commission's findings while modifying part of the decision and setting the fine at approximately €4.125 billion.

The later appeal concerned issues including tying, contractual restrictions, exclusionary effects, Android forks and the assessment of the single and continuous infringement.

Importance for AI-Controlled Corporate Groups

The case illustrates how competition law can examine an interconnected technological ecosystem rather than treating every product or corporate component in isolation.

The same approach may be relevant where AI connects multiple businesses within a corporate group.

Case 6 — Athenian Brewery and Heineken

Case C-393/23, Court of Justice, 2025

This more recent case is important for parent-subsidiary relationships and competition damages litigation.

The Court addressed circumstances involving claims against a parent company and subsidiary and referred to the rebuttable presumption of decisive influence where the parent holds directly or indirectly all or virtually all of the subsidiary's capital.

Later Court case law has expressly summarized this principle.

Importance for AI Groups

Where AI-related conduct is distributed across several companies, questions of corporate control can affect not only regulatory enforcement but also private damages litigation.

Case 7 — Google and Alphabet v Commission (Google Shopping)

Case T-612/17 and subsequent appeal proceedings

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

It is relevant to AI-controlled corporate groups because algorithmic ranking or recommendation systems can influence how competing and affiliated services obtain access to users.

AI Relevance

Imagine a dominant group's AI recommendation engine systematically gives preferential visibility to services belonging to the same corporate group.

Competition analysis may investigate the design, implementation and competitive effects of that arrangement rather than accepting the automated nature of the ranking as a complete explanation.

The Google Shopping proceedings involved Google and Alphabet as group entities and competition allegations concerning digital search services.

15. Key Principles Emerging From the Cases

These cases establish several principles that are highly relevant to AI-controlled corporate groups.

First, legal personality is not decisive. Several corporations can form one undertaking.

Second, economic reality matters. Authorities examine actual control and market behaviour.

Third, parent-company influence matters. Decisive influence can connect liability across a corporate structure.

Fourth, restructuring does not automatically erase responsibility. Competition law follows the relevant economic entity in appropriate circumstances.

Fifth, liability does not automatically extend to every group company. Sumal demonstrates the importance of identifying the relevant economic unit and connection with the infringing activity.

Sixth, digital ecosystems can be examined together. Platform, operating-system, search, application and other interconnected activities may form part of a broader competition analysis.

16. The Difficult Question of Autonomous AI

Future cases may raise a harder issue:

What happens when an AI system makes commercially important decisions with very little immediate human involvement?

Competition law is still fundamentally directed toward undertakings.

The central legal investigation would therefore likely concern matters such as:

Control: Which undertaking controls the system?

Deployment: Who decided to use it?

Objectives: Which commercial objectives were programmed or selected?

Data: Who determines what information enters the model?

Implementation: Which companies execute its decisions?

Benefits: Which undertaking receives the economic benefit?

Supervision: What monitoring and override mechanisms exist?

The fact that an AI system autonomously generated a particular recommendation would not by itself establish that the underlying undertaking falls outside competition law.

17. Hypothetical Example

Consider Alpha Group, consisting of a parent and four subsidiaries:

• Alpha Retail;
• Alpha Logistics;
• Alpha Cloud;
• Alpha Advertising.

The parent introduces AlphaAI, which controls pricing, advertising allocation, logistics access and product recommendations across the group.

AlphaAI begins giving Alpha Retail preferential advertising positions and better logistical conditions while independent retailers receive less favourable treatment.

The legal analysis might examine:

1. whether Alpha Group constitutes one undertaking for the relevant activities;
2. whether it possesses dominance in any relevant market;
3. whether the conduct constitutes exclusionary abuse;
4. whether preferential treatment produces anticompetitive effects;
5. whether independent competitors are being disadvantaged;
6. whether the parent exercised decisive influence;
7. which subsidiaries belong to the relevant economic unit; and
8. whether legitimate objective justifications or efficiencies exist.

The authority would therefore investigate the economic conduct generated through the AI system rather than treating AlphaAI itself as the responsible undertaking.

18. Conclusion

Competition law and AI-controlled corporate groups bring traditional doctrines of corporate responsibility into increasingly automated commercial environments.

The fundamental concepts remain:

undertaking → single economic unit → decisive influence → market power → anticompetitive conduct → responsibility.

AI changes how corporate decisions are generated and coordinated, but it does not automatically change the fundamental competition-law principle that responsibility attaches to the relevant undertaking.

Cases such as ICI, Akzo Nobel, Skanska, Sumal, Google Android, Athenian Brewery/Heineken and Google Shopping provide the existing legal foundation for analysing these issues.

The central future challenge will be determining who exercises economically meaningful control when important commercial decisions are distributed among parent companies, subsidiaries, shared datasets and increasingly autonomous AI systems. Competition authorities will therefore need to distinguish lawful centralized corporate automation from situations where AI facilitates restrictive agreements, exclusionary conduct or the exercise of market power contrary to competition law.

Add a clear legal disclaimerClarify the EU-only scope

Add a clear legal disclaimer

Clarify the EU-only scope

Add an executive summary