Competition Law and Governance of Semantic Market Infrastructures

1. Introduction

Semantic Market Infrastructures (SMIs) are digital or technological infrastructures in which competition depends not merely on the physical or technical ability of systems to connect, but on their ability to understand, classify, interpret and exchange information using common meanings, vocabularies, taxonomies, ontologies, metadata and machine-readable standards.

Examples include:

• AI and machine-learning ecosystems;
• search engines and recommendation systems;
• digital identity and payment systems;
• cloud and data-processing infrastructure;
• interoperability APIs;
• product and service taxonomies;
• financial-data standards;
• health-data interoperability;
• smart-grid and IoT systems;
• e-commerce product classifications;
• digital advertising and measurement systems.

Competition law becomes important because control over the meaning and structure of information can become a source of market power. A dominant undertaking may influence competition by controlling the taxonomy, metadata, API vocabulary, interoperability protocol, classification algorithm, data ontology or semantic standard through which competitors must operate.

The European Commission's current digital-market framework illustrates this evolution. The Digital Markets Act (DMA) expressly addresses interoperability, data access and portability, while the EU Data Act places interoperability standards at the centre of switching and portability between data-processing services.

2. Meaning of Semantic Market Infrastructure

A semantic infrastructure is the information layer that determines what digital objects, transactions, users, products or services mean to a system.

For example, two systems may technically exchange the same data but still be unable to compete effectively if:

System A identifies a product as "electric vehicle battery," while System B's taxonomy divides the same object into "traction battery," "battery pack," "energy-storage module" and "battery management unit."

The problem is therefore not simply connectivity but semantic interoperability.

Main components

1. Taxonomies – hierarchical classification of products, services or users.
2. Ontologies – formal representation of relationships between concepts.
3. Metadata – information describing digital objects.
4. APIs – interfaces allowing systems to communicate.
5. Data schemas – rules determining how information is structured.
6. Identifiers – common references for entities or products.
7. Semantic standards – agreed meanings enabling interoperability.
8. Ranking/classification systems – mechanisms determining how information is interpreted or presented.
9. Machine-readable rules – rules enabling automated decisions.
10. AI models – systems capable of assigning meaning, categorising information and generating outputs.

Thus, semantic infrastructure can become a competitive bottleneck.

3. Competition-Law Significance

Semantic infrastructures create several competition concerns.

A. Control over semantic standards

A dominant firm controlling a widely used taxonomy or ontology may determine:

• which products are recognised;
• which competitors are visible;
• how services are classified;
• which data can be exchanged;
• what constitutes a compatible product.

This can create a form of semantic gatekeeping.

B. Interoperability restrictions

A platform may technically permit access while withholding:

• necessary metadata;
• API documentation;
• semantic mappings;
• authentication information;
• classification rules;
• contextual data.

The result can be formal interoperability without effective interoperability.

C. Semantic self-preferencing

A platform may classify its own products more favourably than competitors.

For example:

Platform-owned service → "premium verified service"

Rival service → "third-party service"

Even where both services satisfy objectively identical criteria, semantic classification can influence rankings and consumer choice.

D. Data portability

Portability is meaningless if users can transfer raw data but cannot transfer the semantic relationships attached to that data.

This is why modern regulatory frameworks increasingly treat interoperability and portability together. The DMA requires real-time portability of user-provided or user-generated data, while the Data Act addresses interoperability standards for data-processing services.

4. Relevant Competition-Law Theories

4.1 Abuse of Dominance

Article 102 TFEU, Section 2 of the Sherman Act, and comparable national provisions can apply where a dominant undertaking uses control over semantic infrastructure to exclude competitors.

Potential abuses include:

• discriminatory access;
• refusal to supply;
• interoperability restrictions;
• tying;
• leveraging;
• self-preferencing;
• discriminatory classification;
• exclusionary technical standards.

5. Essential-Facilities Doctrine

A semantic infrastructure can sometimes resemble an essential facility where competitors cannot realistically compete without access to it.

Traditional essential-facilities jurisprudence emphasises factors such as:

1. indispensability;
2. absence of realistic alternatives;
3. elimination or substantial restriction of competition;
4. objective justification.

However, digital-market cases demonstrate that the strict traditional test does not automatically apply to every interoperability dispute.

The Court of Justice's Android Auto judgment is particularly important because it considered refusal to provide interoperability to a digital platform open to third-party complementors and distinguished such circumstances from the classic Bronner situation.

6. Six Major Case Laws

Case 1 — Commercial Solvents Corp. v Commission

Case: Joined Cases 6/73 and 7/73, Commercial Solvents

Principle

Commercial Solvents established that a dominant undertaking controlling an upstream input cannot necessarily use that control to eliminate competition downstream.

Relevance to semantic infrastructure

The modern analogy is:

Semantic infrastructure → downstream digital service.

If a dominant undertaking controls a critical data vocabulary, identifier, classification system or interoperability layer and then competes downstream using that infrastructure, competition concerns can arise.

Significance

The case establishes an early foundation for analysing leveraging through control of an upstream resource.

Case 2 — RTE and ITP v Commission — Magill

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

Facts

Television broadcasters controlled basic programme information. Magill sought to compile that information into comprehensive television listings.

Principle

The Court recognised circumstances in which refusal to supply information protected by intellectual property could constitute an abuse.

Semantic-market relevance

This is highly relevant because information itself can become an economically important infrastructure.

Modern examples could involve:

• product metadata;
• structured datasets;
• classification information;
• semantic identifiers;
• interoperable content descriptions.

The case demonstrates that control over information can have competition-law consequences where the information is indispensable for a downstream product and refusal produces significant exclusionary effects.

Case 3 — Bronner v Mediaprint

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

Principle

The Court established stringent conditions for imposing an obligation on a dominant undertaking to provide access to infrastructure that it controls.

The traditional analysis focuses on:

• indispensability;
• elimination of effective competition;
• inability to replicate the facility;
• absence of objective justification.

Application to semantic infrastructures

A company controlling a proprietary ontology cannot automatically be required to disclose it simply because competitors would benefit from access.

Competition law must consider:

• whether alternatives exist;
• whether competitors can create substitute semantic systems;
• whether access is technically feasible;
• investment incentives;
• security;
• intellectual-property considerations.

Thus, semantic importance does not automatically equal legal indispensability.

Case 4 — IMS Health v NDC Health

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

Principle

IMS Health concerned a copyrighted pharmaceutical-sales data structure and the refusal to license it.

The case developed the exceptional circumstances under which refusal to license an intellectual-property-protected infrastructure could constitute abuse.

Semantic significance

The famous 1860 brick structure can be understood as an early example of a commercially significant information architecture.

Its competitive importance arose because market participants had organised their business activities around the structure.

Modern analogy

Comparable concerns may arise where an industry becomes dependent upon:

• a dominant product ontology;
• a standardised data architecture;
• proprietary semantic identifiers;
• dominant API schemas.

The more extensively an entire ecosystem coordinates around one semantic architecture, the more significant competition-law questions become.

Case 5 — Microsoft v Commission

Case: T-201/04, Microsoft Corp. v Commission

Principle

Microsoft's refusal to provide interoperability information to competing work-group server products was examined as an exclusionary abuse.

The case is one of the most important authorities concerning interoperability and digital ecosystems.

Semantic-market relevance

Interoperability is not merely the ability to connect systems.

For complex digital systems, competitors need to understand:

• protocols;
• system behaviour;
• communication rules;
• data structures;
• contextual information.

Consequently, Microsoft provides a major foundation for analysing modern semantic infrastructures.

Broader lesson

A dominant undertaking can potentially use control over an interoperability layer to protect its position in an adjacent market.

Case 6 — Google Shopping

Case: Commission Decision AT.39740, Google Search (Shopping), and subsequent EU litigation.

Principle

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

The competition issue involved the relationship between:

• Google's dominant search infrastructure;
• ranking mechanisms;
• visibility;
• competing comparison-shopping services.

EU judicial proceedings examined Google's allegedly unequal positioning and display of competing services.

Semantic-market relevance

Search engines are quintessential semantic infrastructures because they:

1. classify information;
2. interpret queries;
3. associate queries with categories;
4. rank results;
5. determine relevance.

Therefore, ranking and classification can themselves constitute competitive infrastructure.

Case 7 — Google Android

Case: Commission Decision AT.40099, Google Android

Principle

The Android case examined Google's conduct concerning mobile-device ecosystems, including contractual restrictions and leveraging involving Google's search and Android ecosystem.

Semantic relevance

An operating system is not merely software infrastructure.

It determines:

• permissions;
• application access;
• APIs;
• default services;
• data flows;
• interactions between applications.

Consequently, control over the operating-system layer can become control over the semantic and functional relationships between applications.

Case 8 — Alphabet and Others (Android Auto)

Case: C-233/23, Alphabet and Others

This is particularly important for modern semantic-market infrastructure.

The Court of Justice addressed Google's Android Auto ecosystem and interoperability with third-party applications. The judgment of 25 February 2025 clarified that the strict Bronner criteria do not automatically govern every interoperability dispute involving a digital platform that is generally open to third-party complementors.

Importance

The case demonstrates the movement from:

"Must a dominant company surrender its infrastructure?"

towards:

"What interoperability obligations arise from the structure and openness of the digital ecosystem?"

That distinction is crucial for semantic infrastructures.

7. Semantic Self-Preferencing

Semantic infrastructures create a distinctive form of self-preferencing.

A platform may manipulate:

• product categories;
• relevance scores;
• search terminology;
• labels;
• recommendation categories;
• seller classifications;
• trust indicators.

For example:

Platform's own service

"Verified premium provider"

Competitor

"External provider"

The competitive effect may occur before consumers even reach the purchasing stage.

This makes classification power potentially as important as pricing power.

8. Data Access and Semantic Portability

Traditional portability focuses on moving data.

Semantic portability asks a deeper question:

Can the receiving system understand the transferred data in substantially the same way?

For example, transferring:

Customer → Product → Transaction

without transferring the meaning of those relationships may make the data practically unusable.

The DMA now provides end users and authorised third parties with portability rights concerning user-provided or user-generated data, including tools facilitating portability and continuous access.

9. Interoperability Under the DMA

Article 6(7) DMA is especially relevant.

Gatekeepers operating designated operating systems must provide third parties with effective interoperability with relevant hardware and software features controlled by the operating system.

The European Commission has described the purpose as enabling third parties to compete on comparable terms with the gatekeeper's own services.

2026 development

In July 2026, the Commission adopted binding specification measures concerning Google's Android interoperability for AI services.

The measures cover access to Android features necessary for competing AI services, including functions allowing AI assistants to interact with applications and system capabilities.

This is highly significant for semantic markets because AI assistants depend on understanding:

• application context;
• user commands;
• device state;
• underlying data;
• relationships between applications.

10. Search-Data Infrastructure

Semantic competition increasingly depends on access to search-generated data.

In July 2026, the Commission also adopted measures concerning access by third-party search providers to certain Google Search data, including ranking, query, click and view information, subject to conditions concerning anonymisation, access and security.

This illustrates a new regulatory principle:

Competition may require access not merely to a database, but to information generated through the operation of a semantic platform.

11. Semantic Standards and Standardisation

Standardisation can have both pro-competitive and anti-competitive effects.

Pro-competitive effects

Common semantic standards can:

• reduce switching costs;
• facilitate interoperability;
• reduce transaction costs;
• encourage innovation;
• enable multi-homing;
• increase consumer choice.

Anti-competitive risks

A dominant undertaking may:

• manipulate standards;
• exclude competing technologies;
• control standard-setting organisations;
• make interoperability unnecessarily complex;
• introduce proprietary extensions;
• create compatibility asymmetries.

Competition authorities therefore need to examine standard-setting governance.

12. Algorithmic Semantic Governance

AI systems increasingly perform semantic functions automatically.

For example:

Product → category → relevance → recommendation → consumer exposure

An algorithm may determine that two economically substitutable products belong to different categories.

This can influence:

• market visibility;
• search ranking;
• advertising eligibility;
• consumer recommendations;
• platform access;
• pricing opportunities.

Competition authorities may therefore need to investigate not only prices and contracts, but also classification architectures and algorithmic decision rules.

13. Risks of Semantic Lock-In

Semantic lock-in occurs where users or businesses become dependent upon one provider's information architecture.

Examples include:

Vendor lock-in

A cloud provider uses proprietary schemas.

Platform lock-in

Merchants depend upon one platform's product taxonomy.

AI lock-in

Applications are optimised for one AI provider's semantic interface.

Data lock-in

Users cannot transfer meaningful relationships between datasets.

Identity lock-in

A dominant platform controls identifiers necessary for interoperability.

The competitive concern increases where switching requires extensive semantic re-engineering.

14. Competition Concerns in Semantic Market Infrastructures

15. Objective Justifications

Not every restriction concerning semantic infrastructure is anticompetitive.

Legitimate justifications may include:

• cybersecurity;
• privacy;
• data protection;
• intellectual-property protection;
• system integrity;
• prevention of fraud;
• technical limitations;
• safety;
• reliability;
• protection against malicious API use.

The competition-law question is whether the restriction is necessary, proportionate and objectively justified, rather than simply convenient for the dominant undertaking.

The EU's 2026 Android interoperability measures illustrate this balancing exercise: interoperability measures were accompanied by consideration of security and data-protection risks.

16. Remedies

Competition authorities can consider several remedies.

A. Access remedies

Require access to:

• APIs;
• metadata;
• semantic mappings;
• interoperability specifications.

B. Non-discrimination

Require equivalent treatment between:

dominant firm's service ↔ rival service.

C. Data portability

Permit users and businesses to transfer data in usable form.

D. Interoperability

Require technical compatibility across competing systems.

E. Transparency

Require disclosure of:

• ranking criteria;
• classification principles;
• access conditions;
• interoperability specifications.

F. FRAND-style access

Access can sometimes be provided on:

• fair;
• reasonable;
• non-discriminatory

terms.

The DMA's approach to Google's search-data sharing expressly incorporates FRAND-type principles.

17. Governance Framework for Semantic Market Infrastructures

An effective governance model should contain six layers.

Layer 1 — Technical governance

Determine:

• APIs;
• schemas;
• protocols;
• interoperability standards.

Layer 2 — Semantic governance

Determine:

• definitions;
• taxonomies;
• ontologies;
• identifiers.

Layer 3 — Competition governance

Monitor:

• exclusion;
• discrimination;
• self-preferencing;
• foreclosure.

Layer 4 — Data governance

Protect:

• portability;
• privacy;
• security;
• lawful data access.

Layer 5 — Algorithmic governance

Audit:

• classification;
• recommendation;
• ranking;
• automated decision-making.

Layer 6 — Institutional governance

Provide:

• independent oversight;
• complaints mechanisms;
• auditing;
• regulatory intervention.

18. Relationship Between Traditional and Modern Competition Law

The development can be represented as:

Physical infrastructure

↓

Telecommunications infrastructure

↓

Digital infrastructure

↓

Data infrastructure

↓

Interoperability infrastructure

↓

Semantic infrastructure

↓

AI-mediated semantic infrastructure

The competition question consequently moves from:

"Who controls the physical facility?"

to:

"Who controls the technical interface?"

and increasingly to:

"Who controls the meaning through which competing systems interact?"

19. Key Case-Law Principles — Consolidated

20. Conclusion

Semantic Market Infrastructures represent a new layer of competition infrastructure. Their importance lies not merely in storing or transmitting information, but in determining how information is understood, classified, connected and presented.

Traditional competition law already supplies several relevant doctrines:

• abuse of dominance;
• refusal to deal;
• essential facilities;
• tying;
• leveraging;
• discrimination;
• self-preferencing;
• interoperability obligations.

Cases such as Magill, Bronner, IMS Health, Microsoft, Google Shopping and Android Auto demonstrate the gradual movement from physical infrastructure toward information and digital ecosystems.

The principal contemporary challenge is that semantic control can produce market power without looking like traditional infrastructure ownership. A company may possess little physical infrastructure while controlling the taxonomy, metadata, API, ranking architecture, identifiers or ontology upon which an entire ecosystem depends.

Accordingly, effective competition governance increasingly requires regulators to examine three interconnected questions:

Who controls the data?

Who controls interoperability?

Who controls meaning?