Competition Law and Cognitive Data Monopolies

1. Introduction

Cognitive data monopolies refer to situations in which an undertaking obtains, controls, combines, or processes exceptionally valuable datasets that reveal human preferences, behaviour, attention, intentions, decision-making patterns, cognitive states, or inferred characteristics, and that data advantage contributes to durable market power.

The concept is particularly important in the modern digital economy because data can operate simultaneously as:

• an input for artificial intelligence and machine learning;
• a source of consumer profiling;
• an advertising asset;
• a mechanism for improving algorithms;
• a source of network effects;
• a barrier to entry;
• an interoperability resource;
• a source of predictive advantage; and
• a mechanism for reinforcing an existing ecosystem.

Competition law does not generally treat possession of valuable data as unlawful by itself. The legal issue arises when control over data is combined with dominance and conduct that excludes competitors, exploits users, forecloses access, or prevents effective competition.

The European Commission has expressly recognised the competitive importance of access to data. The Digital Markets Act now contains data-access and portability mechanisms, including obligations concerning access to certain search data.

2. Meaning of Cognitive Data

"Cognitive data" is broader than ordinary personal information.

It may include:

A. Direct behavioural data

• searches;
• clicks;
• browsing history;
• purchases;
• viewing history;
• location patterns;
• interaction histories;
• communications metadata.

B. Inferred cognitive information

Platforms may use behavioural data to infer:

• preferences;
• interests;
• purchasing intentions;
• political or commercial interests;
• vulnerabilities;
• likely future behaviour;
• consumer segments;
• attention patterns.

C. AI-generated cognitive datasets

Modern AI systems can generate proprietary datasets from:

• user prompts;
• queries;
• feedback;
• corrections;
• conversations;
• model interactions;
• click-through behaviour;
• ranking responses;
• reinforcement signals.

D. Ecosystem data

A vertically integrated technology company may combine data from:

Search → browser → operating system → app store → payments → maps → advertising → cloud → AI

This creates the possibility of a data feedback loop.

3. The Cognitive Data Monopoly Feedback Loop

A typical cognitive-data ecosystem can operate as follows:

More users

↓

More behavioural data

↓

More accurate profiles and predictions

↓

Better algorithms / AI

↓

Better products

↓

More users

↓

More data

This is sometimes described as a data-network-effect or data-feedback loop.

The competitive concern is particularly serious where competitors cannot reproduce the relevant dataset even if they possess comparable technology.

The Bundeskartellamt has expressly recognised that collection, processing and combination of data can reinforce the market power of large digital companies.

4. When Does Cognitive Data Become a Competition Problem?

A competition-law concern normally requires more than simply possessing data.

The principal questions are:

1. Is the undertaking dominant?
2. Is the data strategically important to competition?
3. Can competitors realistically reproduce the data?
4. Does the undertaking restrict access to the data?
5. Does it combine data across markets or services?
6. Does it use data to foreclose competitors?
7. Does it use data to advantage its own downstream service?
8. Does it impose unfair or exploitative data terms?
9. Does it use acquisitions to eliminate emerging data competitors?
10. Does the conduct produce durable entry barriers?

5. Relevant Competition-Law Theories

A. Abuse of Dominance

Under Article 102 TFEU and comparable national provisions, cognitive-data conduct may constitute abuse where a dominant undertaking uses its position to:

• exclude competitors;
• impose unfair conditions;
• deny indispensable access;
• tie services;
• discriminate between data users;
• engage in self-preferencing;
• restrict interoperability.

B. Refusal to Supply / Access to Data

A dominant undertaking may possess a dataset that competitors cannot reasonably reproduce.

A refusal to provide access becomes legally significant where the stringent conditions developed under the essential-facilities / refusal-to-supply doctrine are satisfied.

Important considerations include:

• indispensability;
• lack of realistic substitutes;
• elimination of effective competition;
• objective justification;
• whether access is technically feasible.

C. Data Leveraging

A company dominant in one market may use its data advantage to enter or strengthen its position in another market.

For example:

Dominant search engine

→ collects search behaviour

→ develops superior advertising profiles

→ strengthens advertising platform

→ generates more revenue

→ finances further technological development.

This can raise leveraging concerns.

6. Data Combination as an Antitrust Issue

One of the most important developments is the recognition that combining datasets may itself have competitive significance.

Suppose a platform separately possesses:

• search data;
• social-network data;
• messaging data;
• shopping data;
• location data.

Individually, each dataset may have competitors or substitutes.

But combining them may generate a much richer behavioural profile.

The resulting competitive advantage can therefore be substantially greater than the sum of the individual datasets.

7. Major Case Laws

1. Meta Platforms Inc. v Bundeskartellamt — C-252/21

Court: Court of Justice of the European Union
Year: 2023

This is one of the most important cases concerning data and competition law.

The German Bundeskartellamt challenged Meta's practice of combining data obtained from Facebook with information from other Meta services and third-party websites and applications.

The CJEU held that a competition authority can, when examining abuse of dominance, consider whether data processing complies with the GDPR, while respecting the institutional responsibilities of data-protection authorities.

Competition-law significance

The case demonstrates that:

• personal-data processing can have competition implications;
• privacy and competition law can overlap;
• data accumulation may contribute to market power;
• exploitative terms imposed by a dominant digital platform can be scrutinised;
• data protection cannot automatically be separated from competition analysis.

The case is particularly important for cognitive data because combining datasets can permit highly detailed inferences concerning users' preferences and behaviour.

2. Bundeskartellamt Facebook Data-Combination Decision

Authority: German Bundeskartellamt
Year: 2019

The Bundeskartellamt found that Facebook had abused its dominant position by making use of its social network conditional on extensive collection and combination of data from Facebook, Instagram, WhatsApp and third-party sources without genuinely voluntary consent.

The authority considered the extensive collection, combination and exploitation of data to be an exploitative abuse and also identified competitive disadvantages for rivals unable to accumulate a comparable data resource.

Significance

This case established an important analytical proposition:

Data accumulation can itself reinforce dominance where the dataset improves the competitive position of the dominant platform.

It is therefore a foundational authority for the concept of a cognitive-data monopoly.

3. Google Shopping — Google LLC and Alphabet v Commission, C-48/22 P

Court: CJEU
Judgment: 10 September 2024

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

The CJEU upheld the competition-law finding concerning Google's conduct, addressing the distinction between competition on the merits and conduct capable of foreclosing competitors.

Relevance to cognitive data

Search engines generate enormous quantities of behavioural signals:

• queries;
• clicks;
• rankings;
• interactions;
• product interests;
• consumer intent.

Preferential treatment of the dominant platform's own services can therefore affect both visibility and accumulation of commercially valuable behavioural information.

The case illustrates how a platform's control over an important information gateway can reinforce ecosystem advantages.

4. IMS Health GmbH & Co. OHG v NDC Health — C-418/01

Court: CJEU
Year: 2004

IMS Health controlled a commercially important "brick structure" used in supplying pharmaceutical sales data.

The CJEU considered the circumstances in which refusal by a dominant undertaking to provide access to an indispensable resource can constitute abuse. The Court emphasised the importance of determining whether the resource was indispensable and whether refusal could eliminate effective competition in a downstream market.

Significance for cognitive data

IMS Health provides a useful analytical foundation for modern data-access disputes.

A proprietary dataset may potentially become competition-law significant when:

• competitors cannot reasonably reproduce it;
• it is indispensable for competing;
• access is objectively necessary;
• refusal eliminates effective competition.

Thus, the case is highly relevant to AI training datasets, behavioural databases and specialised predictive datasets.

5. Magill — RTE and ITP v Commission, Joined Cases C-241/91 P and C-242/91 P

Court: CJEU
Year: 1995

The case concerned television programme information controlled through copyright.

The CJEU developed the exceptional circumstances under which refusal to license intellectual property by a dominant undertaking can constitute abuse.

The case involved a dominant control over important information and the possibility of preventing the emergence of a new product for which there was consumer demand.

Relevance to cognitive-data monopolies

The underlying principle can be relevant where:

Proprietary information resource → downstream dependence → exclusion of innovative competitors

For example, analogous issues could arise with:

• specialised datasets;
• proprietary behavioural information;
• unique AI training resources;
• structured databases;
• information necessary for downstream predictive services.

The threshold remains exceptionally high; possession of protected information does not automatically create an obligation to license it.

6. Microsoft — Commission Decision concerning interoperability

Authority: European Commission
Year: 2004

The Microsoft case involved Microsoft's refusal to disclose interoperability information and the tying of Windows with Windows Media Player.

The Commission required Microsoft to disclose relevant interoperability information under specified conditions and imposed a separate remedy concerning the tying conduct.

Relevance to cognitive data

The case illustrates an important principle:

Control over a technological interface can become a competition problem when rivals depend upon it to compete effectively.

In contemporary markets, the analogous resource may be:

• API access;
• interoperability data;
• user-generated data;
• platform telemetry;
• model interaction data;
• technical metadata.

Therefore, cognitive-data monopolies may involve not merely ownership of data but control over the interfaces through which competitors can obtain or use data.

7. United States v Google — Search and Search Advertising

Court: U.S. District Court for the District of Columbia
Key judgment: 2024; remedies developed subsequently

The U.S. case concerning Google Search found Google liable under Section 2 of the Sherman Act for maintaining monopoly power through exclusionary distribution agreements.

The court's findings included the importance of user data to improving search quality and the competitive importance of achieving sufficient scale.

Subsequent remedies ordered Google to make specified search-index and user-interaction data available to certain competitors and potential competitors.

Significance

This is particularly important for the cognitive-data monopoly concept.

It demonstrates the possible relationship:

Scale → queries → user interaction data → improved search → more users → greater scale.

The remedy therefore treats certain categories of search data as potentially important to restoring competitive conditions.

8. FTC v Facebook / Meta

Authority: U.S. Federal Trade Commission
Proceeding: Section 2 monopolisation litigation

The FTC alleged that Facebook maintained monopoly power in personal social networking through acquisitions and conduct concerning API access, including the acquisitions of Instagram and WhatsApp. The case remains significant as an illustration of the relationship between data-rich ecosystems, acquisitions and potential elimination of emerging competitive threats.

Significance

Data monopolisation can occur not only through conduct involving existing datasets but also through:

Acquisition → elimination of emerging rival → consolidation of user base/data → strengthening of ecosystem

This is especially important in AI and cognitive-technology markets where a small emerging platform may possess strategically important interaction data.

8. Comparative Case-Law Principles

9. Cognitive Data as an Essential Facility

The argument that a cognitive dataset is an essential facility must be treated cautiously.

A dataset is more likely to attract competition-law scrutiny when it is:

1. Indispensable

Competitors cannot reasonably reproduce the dataset.

2. Unique

The data possess characteristics that alternative datasets cannot replicate.

3. Continuously generated

The incumbent receives new data every day while competitors cannot obtain equivalent flows.

4. Necessary for downstream competition

The dataset materially affects the ability to compete.

5. Controlled by a dominant undertaking

The resource is controlled by an undertaking possessing substantial market power.

6. Refused without objective justification

A refusal cannot be justified by legitimate security, privacy, intellectual-property or technical considerations.

10. Data Portability

Data portability can reduce cognitive-data monopolisation.

Traditional portability focuses on enabling users to take their information elsewhere.

In platform competition, broader data access may also concern:

• business-user data;
• search data;
• advertising-performance data;
• interoperability information;
• platform-generated data.

The EU's DMA expressly contains data-access mechanisms designed to give businesses and users greater access to valuable data.

11. Interoperability

Interoperability is closely connected to data competition.

Without interoperability:

Platform A

→ controls data

→ prevents transfer

→ users remain on A

→ competitors cannot obtain equivalent data

→ network effects strengthen A.

With interoperability:

Platform A ↔ Platform B ↔ Platform C

Data can move or interact under legally permitted conditions.

This can reduce switching costs and make entry easier.

The Microsoft interoperability case provides an important historical foundation for this analysis.

12. Self-Preferencing and Cognitive Data

A dominant platform may use its data advantage to favour its own products.

Example:

Search platform

collects:

• consumer queries;
• clicks;
• purchase intentions.

It then launches its own:

shopping service

and uses its information advantage to improve ranking, targeting and product design.

The competitive concern is not merely that the company possesses data, but that control of the data is combined with preferential treatment of its downstream service.

Google Shopping provides an important framework for analysing this type of conduct.

13. AI and Cognitive Data Monopolies

The issue becomes particularly significant with generative AI.

AI firms may possess:

• enormous prompt datasets;
• human feedback;
• preference data;
• search queries;
• coding interactions;
• multimodal inputs;
• reinforcement signals;
• model evaluation data.

These can create a learning advantage.

Data feedback loop in AI

More users

↓

More prompts and feedback

↓

More training/evaluation data

↓

Better model

↓

More users

This can produce significant barriers to entry.

Competition authorities therefore may increasingly examine whether AI companies:

• restrict data portability;
• prevent interoperability;
• acquire data-rich startups;
• impose exclusivity;
• tie AI models to dominant ecosystems;
• discriminate in API access;
• exploit proprietary datasets;
• use customer data to disadvantage downstream competitors.

14. Cognitive Data and Merger Control

Data monopolisation can also develop through acquisitions.

A merger may combine:

Company A's behavioural data

Company B's cognitive/inference data

=

Combined predictive dataset

The combined dataset may provide advantages unavailable to competitors.

Merger authorities can therefore examine:

• data overlap;
• data substitutability;
• privacy dimensions;
• network effects;
• entry barriers;
• potential competition;
• innovation effects;
• future data accumulation.

The FTC's Meta/Within proceeding illustrates how digital acquisitions may be examined for effects on innovation and emerging technology markets.

15. Exploitative Abuse

Cognitive-data monopolies can produce consumer exploitation even when the service is nominally free.

The economic exchange may be:

Free service

↕

Consumer attention + behavioural data + inferred preferences

The competition-law question becomes whether a dominant undertaking imposes unfair conditions concerning the collection or exploitation of this data.

The German Facebook proceedings are particularly important here.

16. Data as a Barrier to Entry

A cognitive-data incumbent can possess advantages that are difficult for new entrants to reproduce.

17. Possible Antitrust Remedies

Competition authorities may consider several remedies.

Structural remedies

• divestiture;
• separation of businesses;
• restrictions on acquisitions.

Behavioural remedies

• data access;
• interoperability;
• non-discrimination;
• API access;
• prohibition of self-preferencing;
• restrictions on data combination.

Consumer-facing remedies

• meaningful consent;
• data portability;
• switching mechanisms;
• transparency.

Data remedies

• FRAND-type access;
• anonymised data sharing;
• real-time data access;
• data silos;
• restrictions on cross-service combination.

The appropriate remedy depends on the particular market failure; compelled access is not automatically justified merely because a dataset is valuable.

18. Privacy and Competition Law Must Be Coordinated

Cognitive-data monopolies demonstrate that competition law and privacy law can overlap without becoming identical.

Competition law asks:

Does the conduct harm the competitive process?

Privacy law asks:

Is personal-data processing lawful and appropriately controlled?

The Meta judgment illustrates that these legal regimes can interact in a dominance analysis, while also emphasising coordination with data-protection authorities.

19. Contemporary Regulatory Development

The European Union's DMA represents an important movement from purely ex-post antitrust enforcement toward ex-ante regulation of gatekeepers.

The Commission has specifically required Google to provide certain search data to eligible competitors under Article 6(11), with measures concerning query, ranking, click and view information.

The Commission's 2026 decision on implementation of these measures demonstrates how access to strategically valuable search data is increasingly being treated as a competition-policy issue.

This is highly relevant to cognitive data because search interactions can reveal consumer intent and preferences, rather than merely static personal information.

20. Key Legal Tests

A useful analytical framework for cognitive-data monopoly cases is:

Step 1 — Define the relevant market

Identify whether the relevant market concerns:

• search;
• social networking;
• digital advertising;
• AI services;
• data brokerage;
• specialised datasets;
• cloud services;
• app ecosystems.

Step 2 — Establish dominance

Examine:

• market shares;
• entry barriers;
• network effects;
• switching costs;
• data advantages;
• ecosystem power.

Step 3 — Identify the data advantage

Ask:

• What data does the undertaking possess?
• Is it unique?
• Is it continuously generated?
• Can competitors replicate it?

Step 4 — Identify the conduct

Possible conduct includes:

• refusal to provide access;
• discriminatory access;
• self-preferencing;
• tying;
• exclusive arrangements;
• data combination;
• exploitative terms;
• acquisitions.

Step 5 — Establish competitive effects

Examine:

• foreclosure;
• entry barriers;
• reduced innovation;
• reduced consumer choice;
• higher advertising costs;
• reduced quality;
• diminished privacy;
• reduced technological diversity.

Step 6 — Consider objective justification

Possible justifications include:

• privacy;
• cybersecurity;
• intellectual property;
• technical feasibility;
• legitimate commercial confidentiality;
• data security.

Step 7 — Select proportionate remedy

Possible remedies:

Access → interoperability → non-discrimination → portability → structural separation

depending on the infringement.

21. Important Distinction: Data Monopoly vs Data Advantage

It is important not to equate:

large dataset = monopoly

or

valuable data = essential facility.

Competition law generally requires an assessment of market power, conduct and competitive effects.

A company may possess an enormous dataset while facing vigorous competition.

Conversely, a comparatively smaller dataset can be strategically decisive if it is unique, indispensable and continuously generated.

The critical question is therefore:

Does control over cognitive data create or reinforce market power in a way that the undertaking uses to restrict competition or exploit users?

22. Emerging Issues

Future litigation is likely to concern:

A. AI training data

Whether dominant firms can exclusively control datasets necessary for competing AI models.

B. Synthetic data

Whether synthetic data can genuinely substitute for real behavioural data.

C. Cognitive inference

Whether competition law should consider not merely collected data but information inferred from data.

D. Real-time behavioural data

Whether continuous data streams create stronger entry barriers than historical datasets.

E. Data feedback loops

Whether the combination of network effects and machine learning produces self-reinforcing dominance.

F. Data interoperability

Whether competitors should receive access to particular categories of platform-generated information.

G. Data combination

Whether combining otherwise separate datasets constitutes an exclusionary or exploitative practice.

H. AI acquisitions

Whether acquisition of a small company with an unusually valuable dataset eliminates future competition.

23. Conclusion

Cognitive data monopolies represent a modern form of competition concern in which market power can arise not simply from ownership of physical assets or conventional intellectual property, but from control over behavioural information, inferred preferences, interaction data and continuously generated learning signals.

The principal competition-law concerns include:

1. data-driven dominance;
2. data combination;
3. refusal of access to indispensable datasets;
4. self-preferencing;
5. data-based leveraging;
6. interoperability restrictions;
7. exploitative data terms;
8. data-driven network effects;
9. data-rich acquisitions; and
10. AI learning feedback loops.

The case law from Meta, Facebook, Google Shopping, IMS Health, Magill, Microsoft and U.S. Google litigation demonstrates that competition law already possesses several doctrines capable of addressing aspects of data-driven market power.

The emerging regulatory direction is increasingly toward combining traditional abuse-of-dominance and merger-control principles with data access, portability, interoperability and gatekeeper obligations. The central legal challenge will be to distinguish legitimate data-driven innovation from situations in which control over cognitive data becomes a durable mechanism for excluding competitors and entrenching market power.