Competition Law and Knowledge Graph Ownership and Market Power

1. Introduction

A knowledge graph is a structured database that represents entities—such as people, companies, products, locations, concepts and events—and the relationships between them. Search engines, AI systems, digital assistants, recommendation platforms and enterprise software may use knowledge graphs to improve search results, answer questions, identify entities and generate contextual information.

From a competition-law perspective, the important issue is not simply who owns the knowledge graph, but whether control over a commercially important knowledge graph creates or reinforces market power, and whether that control is used to exclude competitors.

A knowledge graph can become competitively significant because it may combine:

• proprietary information;
• publicly available information;
• user-generated information;
• search and clickstream data;
• entity-resolution technology;
• relationship data;
• location and business information;
• structured metadata;
• machine-learning outputs; and
• continuously updated proprietary datasets.

The Competition Commission of India has recognised more generally that control over data can contribute to market power in digital markets, particularly where network effects and data accumulation reinforce one another.

2. Meaning of Knowledge Graph Ownership

Ownership or control may arise through several mechanisms.

A. Intellectual-property ownership

A company may claim copyright, database rights, trade-secret protection or contractual rights over portions of a knowledge graph.

B. Control over underlying data

Even where individual facts cannot ordinarily be monopolised, a company may control:

• the database containing those facts;
• proprietary data-cleaning processes;
• entity-matching systems;
• data collection pipelines;
• APIs;
• metadata;
• historical datasets; and
• continuously updated information.

C. Control over the technology

The competitive asset may be the software that determines how entities and relationships are identified.

For example:

Person A → CEO of → Company B → headquartered in → City C → subsidiary of → Company D.

The individual facts may be available elsewhere, but a highly comprehensive and accurate system connecting them may be difficult for competitors to reproduce.

D. Contractual control

A platform may obtain exclusive or preferential rights to data from:

• publishers;
• retailers;
• mapping providers;
• content creators;
• application developers;
• business directories; and
• other data suppliers.

This can raise vertical-exclusion concerns.

3. Relevant Competition-Law Framework

Knowledge-graph ownership can potentially implicate several areas of competition law.

Section 3 – Anti-competitive agreements

In India, arrangements concerning a knowledge graph may raise Section 3 issues where agreements between enterprises:

• restrict access to essential information;
• impose exclusivity;
• prevent interoperability;
• restrict data portability;
• allocate data sources; or
• foreclose competing information providers.

Section 4 – Abuse of dominant position

Where the owner possesses substantial market power, potential concerns include:

• refusal to provide access;
• discriminatory access;
• tying;
• bundling;
• self-preferencing;
• leveraging;
• exclusionary licensing;
• excessive contractual restrictions;
• discriminatory API access; and
• denial of market access.

Section 5 – Combinations

Acquisition of another company's:

• knowledge graph;
• structured database;
• entity-resolution technology;
• proprietary data repository; or
• data-rich platform

may create competition concerns even where the acquired company has relatively little conventional revenue.

Data as a source of market power

The important distinction is between data ownership and competitive control.

Ownership alone does not automatically establish dominance. Competition authorities would normally examine factors such as:

1. uniqueness of the data;
2. accuracy;
3. comprehensiveness;
4. frequency of updating;
5. substitutability;
6. cost of replication;
7. network effects;
8. switching costs;
9. interoperability;
10. access to alternative data sources; and
11. whether competitors can obtain comparable information within a commercially reasonable period.

4. When Can a Knowledge Graph Create Market Power?

A knowledge graph becomes particularly important when it produces a data feedback loop:

More users → more queries → more behavioural information → better graph → better results → more users

This can produce cumulative advantages.

The competitive significance therefore may extend beyond the database itself.

For example:

Search engine → knowledge graph → better search answers → more users → more queries → additional data → improved knowledge graph.

A competitor with a smaller user base may face difficulty reproducing the same quality even if it possesses adequate computing resources.

5. Knowledge Graph as a Potential Essential Input

The most difficult issue is whether a knowledge graph can constitute an essential facility or indispensable input.

Competition law generally does not require a dominant undertaking to provide every proprietary asset to competitors.

A refusal becomes more problematic where the input is:

• objectively necessary;
• practically impossible or economically unreasonable to reproduce;
• indispensable for effective competition;
• controlled by a dominant undertaking; and
• capable of being supplied without eliminating legitimate incentives to innovate.

This makes essential-facility doctrine particularly relevant.

6. Six Important Case Laws

1. Magill TV Guide/ITP, BBC and RTÉ

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

Principle

The European Court of Justice developed the exceptional circumstances surrounding compulsory access to intellectual property.

The case concerned television-programme information that was controlled by broadcasters and needed by publishers of comprehensive television guides.

The Court identified circumstances in which refusal to license an intellectual-property-protected resource could constitute abuse of dominance.

Relevance to knowledge graphs

A knowledge graph owner could potentially argue that its database is proprietary intellectual property.

However, Magill demonstrates that intellectual-property protection does not necessarily provide an absolute competition-law immunity.

For a knowledge graph, the critical questions would include:

• Is the information indispensable?
• Is there a genuine substitute?
• Does refusal prevent the emergence of a new product?
• Is the refusal capable of excluding effective competition?

The threshold is exceptional and should not be treated as an automatic access right.

2. IMS Health v NDC Health

Case: C-418/01, IMS Health GmbH & Co. OHG v NDC Health GmbH & Co. KG.

Principle

The Court applied the exceptional-refusal framework to a pharmaceutical sales-information system.

The dispute concerned the structure of pharmaceutical sales data and the difficulty competitors faced in reproducing the relevant system.

Relevance

This case is particularly important for knowledge graphs because it illustrates the significance of replicability.

A knowledge graph may become competitively important if competitors cannot realistically recreate:

• its ontology;
• entity structure;
• historical relationships;
• data accuracy;
• geographic coverage; and
• continuously updated connections.

However, the existence of substantial replication costs does not automatically mean that compulsory access is required.

The exceptional conditions must still be satisfied.

3. Bronner v Mediaprint

Case: C-7/97, Oscar Bronner GmbH & Co. KG v Mediaprint.

Principle

The Court adopted a strict approach to refusal-to-deal claims.

The fact that access to another undertaking's infrastructure would make competition easier is not sufficient.

The facility must generally be indispensable, with no realistic alternative.

Application to knowledge graphs

Suppose a dominant search platform refuses access to its proprietary knowledge graph.

A competitor would need to establish more than:

"Access would make our search engine better."

The relevant question would be whether the competitor can realistically build or obtain an alternative graph.

If alternative sources exist—including public databases, licensing arrangements, crowdsourced information or independent data collection—the Bronner standard becomes more difficult to satisfy.

4. Google Shopping

Case: Google Search (Shopping), European Commission Decision AT.39740; General Court, Case T-612/17.

Principle

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

The case demonstrated how control over a major digital platform can provide an opportunity to favour an affiliated service.

The General Court largely upheld the Commission's findings concerning Google's conduct in 2021.

Knowledge-graph relevance

The same structural problem can arise with knowledge graphs.

Suppose a search engine controls a highly valuable knowledge graph and uses that information to:

• provide its own answer service;
• rank its own products;
• populate its own commercial services;
• enhance its own AI assistant; or
• disadvantage rival information services.

The competition issue would therefore not necessarily be ownership of the graph, but leveraging control over the graph into an adjacent market.

5. Google Android

Case: Google LLC and Alphabet Inc. v European Commission, Case T-604/18.

Principle

The EU General Court examined Google's contractual arrangements involving Android, Google Play, Search, Chrome and anti-fragmentation restrictions.

The Court treated the conduct within the broader context of Google's position across interconnected digital markets.

Relevance

Knowledge graphs frequently operate within ecosystems.

A company might combine:

Search + Knowledge Graph + Browser + Maps + Assistant + AI + Advertising.

If control of one component reinforces another component's market position, competition authorities may examine ecosystem leveraging rather than treating every product as an isolated market.

The Indian CCI has similarly examined Google's Android ecosystem through relevant markets involving licensable mobile operating systems and app stores, including questions of leveraging and market access.

6. Google Search Data / Knowledge Graph Remedies in the United States

The recent U.S. Google search remedies litigation is especially relevant because proposed remedies have specifically discussed Google's Knowledge Graph.

In the 2025 remedies proceedings, plaintiffs proposed requiring disclosure of databases containing information sufficient to recreate Google's Knowledge Graph, including local information. The court described the Knowledge Graph as a database containing information concerning people, places and things and the relationships connecting them.

Competition significance

This is highly significant conceptually.

It demonstrates that competition authorities and courts can view a knowledge graph not merely as an intellectual-property asset but as a potentially important competitive input.

The remedy question becomes:

Can controlled access to knowledge-graph information reduce structural barriers faced by competing search or AI services?

This is different from saying that every competitor is entitled to the complete graph.

The appropriate remedy may instead concern:

• specified categories of information;
• APIs;
• periodic data access;
• interoperability;
• non-discriminatory access;
• data portability; or
• limited disclosure sufficient to facilitate competition.

7. Additional Relevant Case: Google Android – India

The Indian Google Android proceedings provide an important domestic analogy.

The CCI found Google dominant in relevant markets involving licensable mobile operating systems and Android app stores and examined Google's contractual arrangements with device manufacturers. It concluded that certain arrangements involving pre-installation, Android forks and market access infringed Section 4.

The significance for knowledge graphs is the leveraging principle.

A company possessing dominance in one digital layer cannot necessarily use that position to disadvantage competitors in related markets.

Thus:

Dominant search/data infrastructure
↓
Control over knowledge graph
↓
Preferential treatment of own AI/search product
↓
Reduced access for rival services

could potentially raise Section 4 concerns depending on the relevant market and evidence.

8. Self-Preferencing and Knowledge Graphs

Self-preferencing is potentially significant because the graph owner may control both:

1. the underlying information infrastructure; and
2. the consumer-facing service.

For example, a platform could potentially give its own service:

• richer entity descriptions;
• earlier access to new relationships;
• greater data granularity;
• preferential API access;
• superior geographic information;
• enhanced structured results; or
• preferential ranking.

The competition question is whether this conduct produces exclusionary effects rather than simply reflecting legitimate product improvement.

9. Refusal to License the Knowledge Graph

A refusal to license can fall into several categories.

10. Knowledge Graphs and Merger Control

Knowledge-graph acquisitions can create competition concerns even where traditional turnover thresholds appear modest.

Consider:

Company A: dominant search engine
Company B: specialist medical knowledge graph

The transaction could eliminate an emerging competitor or give Company A exclusive access to highly valuable structured medical information.

Potential theories include:

A. Horizontal effects

Two competing knowledge-graph providers combine.

B. Vertical effects

A search platform acquires an upstream information supplier.

C. Conglomerate effects

A large digital ecosystem acquires a graph that strengthens several adjacent services.

D. Data concentration

The transaction combines separate datasets and produces a substantially stronger information asset.

11. Knowledge Graphs and AI Competition

The issue has become particularly important with generative AI.

AI systems may require structured information concerning:

• entities;
• relationships;
• factual verification;
• geographic information;
• commercial information;
• scientific relationships;
• product attributes; and
• real-world events.

A dominant knowledge-graph owner could therefore potentially gain advantages in AI services.

The competitive feedback loop may become:

Knowledge graph → AI model/service → users → queries → data → improved graph → improved AI

This creates a possible cross-market data advantage.

12. Interoperability as a Competition Remedy

Where competition concerns are established, regulators may consider remedies such as:

1. API access

Competitors receive controlled technical access.

2. Data portability

Users or businesses can transfer relevant information.

3. Interoperability

The knowledge graph can interact with competing systems.

4. FRAND-style access

Access may be offered on fair, reasonable and non-discriminatory terms where appropriate.

5. Non-discrimination

The owner cannot provide substantially better technical access to its own downstream service than to similarly situated competitors.

6. Data-sharing remedies

Specific datasets may be made available without requiring disclosure of the entire proprietary system.

The EU's Digital Markets Act provides a contemporary example of competition regulation addressing access to valuable data: the Commission's 2026 proceedings concerning Google Search data required sharing of anonymised search data with eligible competing search engines under fair, reasonable and non-discriminatory terms.

13. Limits of Competition-Law Intervention

Knowledge-graph ownership should not automatically be equated with monopoly power.

Several safeguards are important.

First

Facts themselves may not be capable of exclusive ownership in the same way as a creative database structure or proprietary software.

Second

Competition law generally protects competition rather than guaranteeing competitors access to every successful firm's assets.

Third

Mandatory access can reduce incentives to invest in:

• data collection;
• verification;
• infrastructure;
• ontology development;
• entity resolution; and
• innovation.

Fourth

A competitor's difficulty in reproducing a graph does not automatically establish an antitrust violation.

The Bronner and IMS Health lines of authority illustrate why indispensability and exceptional circumstances matter.

14. Key Legal Tests

For examination purposes, the issue can be analysed through the following framework:

Step 1 – Define the relevant market

Possible markets may include:

• general search;
• specialised search;
• knowledge services;
• structured information services;
• mapping;
• AI assistants;
• enterprise data services; or
• data/API services.

Step 2 – Determine market power

Examine:

• market share;
• barriers to entry;
• network effects;
• data advantages;
• switching costs;
• interoperability;
• economies of scale; and
• access to alternative datasets.

Step 3 – Identify the controlled asset

Determine whether the undertaking controls:

• the graph itself;
• underlying data;
• APIs;
• ontology;
• entity-resolution technology;
• update mechanisms; or
• complementary infrastructure.

Step 4 – Examine the conduct

Possible conduct includes:

• refusal to deal;
• discriminatory access;
• self-preferencing;
• tying;
• exclusive dealing;
• bundling;
• data foreclosure;
• interoperability restrictions; or
• leveraging.

Step 5 – Assess effects

Ask whether the conduct:

• excludes rivals;
• raises rivals' costs;
• prevents entry;
• reduces innovation;
• limits consumer choice;
• increases switching costs; or
• protects dominance in an adjacent market.

Step 6 – Examine objective justification

The undertaking may rely on:

• privacy;
• cybersecurity;
• intellectual-property protection;
• data accuracy;
• technical integrity;
• security;
• legitimate commercial investment; or
• protection against free-riding.

15. Six-Case-Law Summary

16. Conclusion

Knowledge-graph ownership is not inherently anti-competitive. The central competition-law question is whether control over the graph creates a significant competitive advantage and whether the owner uses that advantage in an exclusionary manner.

The strongest competition concerns arise where a knowledge graph is:

unique + difficult to replicate + commercially indispensable + controlled by a dominant undertaking + used to disadvantage competing services.

The legal analysis therefore moves from ownership → market power → indispensability → conduct → foreclosure → effects → justification → remedy.

The Magill, IMS Health and Bronner cases establish the restrictive framework for compulsory access, while Google Shopping, Google Android, and the more recent Google search remedies proceedings demonstrate how data, platform ecosystems and structured information can become central to modern competition-law analysis. The contemporary regulatory direction also shows increasing attention to data access, interoperability and non-discriminatory sharing as possible tools for addressing entrenched digital market power.