Competition Law And Competition Governance In Cognitive Economies .
Competition Law and Competition Governance in Cognitive Economies
1. Introduction
A cognitive economy is an economy in which economic decisions are increasingly made, assisted, or optimized by artificial intelligence, machine learning, algorithms, automated agents, predictive analytics, behavioural profiling, and large-scale data systems.
In such an economy, competition is no longer determined only by traditional factors such as price, output, physical assets, and market share. Competitive power may instead arise from control over:
- data and datasets;
- AI models and computing infrastructure;
- algorithms and recommendation systems;
- consumer attention and behavioural information;
- digital ecosystems;
- cloud and semiconductor infrastructure;
- application programming interfaces (APIs);
- interoperability;
- default settings;
- automated decision-making;
- network effects;
- switching costs; and
- continuously learning systems.
Competition governance therefore has to address a fundamental question:
How should competition law operate when the economically significant decisions affecting markets are increasingly made by machines and data-driven systems?
Existing competition law remains applicable, but its application must account for the special characteristics of cognitive markets.
2. Meaning of a Cognitive Economy
A cognitive economy may be understood as an economic system in which information-processing capacity becomes a principal source of competitive advantage.
Traditional economy:
Capital → Production → Distribution → Consumer
Cognitive economy:
Data → Algorithm → Prediction → Automated decision → Behavioural feedback → More data → Improved algorithm
This produces a potentially self-reinforcing competitive cycle.
Example
An AI-powered platform may:
- collect millions of consumer interactions;
- analyse purchasing behaviour;
- predict consumer preferences;
- rank products according to those predictions;
- observe which recommendations succeed;
- obtain additional behavioural data; and
- continuously improve its model.
The resulting advantage may become difficult for competitors to reproduce.
This creates competition-law questions concerning data concentration, algorithmic discrimination, self-preferencing, interoperability, exclusion, collusion and innovation foreclosure.
3. Main Competition-Law Characteristics
A. Data as a competitive resource
Data may function as an important input into AI systems.
A dominant undertaking may possess:
- proprietary datasets;
- search histories;
- transaction data;
- location data;
- behavioural profiles;
- training data;
- customer interaction data; and
- feedback generated by AI systems.
However, possession of large amounts of data is not automatically unlawful. Competition law generally becomes relevant when control over data contributes to exclusionary conduct, discriminatory access, tying, refusal to supply, exploitative conduct, or barriers to entry.
4. Algorithmic Market Power
Market power in cognitive economies may arise without conventional price increases.
A platform may instead control:
- visibility;
- ranking;
- recommendation;
- access to consumers;
- advertising allocation;
- search results;
- transaction opportunities.
Consequently, competition authorities may have to examine non-price dimensions of competition.
The European Commission's current Digital Markets Act framework illustrates this development. In 2026, the Commission has specifically addressed access to Google's search data and interoperability between competing AI services and Android.
5. Network Effects and Cognitive Feedback Loops
AI markets can exhibit particularly strong network effects.
A simplified model is:
More users → More data → Better AI → Better service → More users
This creates a feedback loop.
A large incumbent may therefore possess an advantage that is not simply based on its existing market share but on its ability to continuously improve its system through user interactions.
Competition analysis may consequently consider:
- direct network effects;
- indirect network effects;
- data network effects;
- learning effects;
- switching costs;
- economies of scale;
- economies of scope; and
- feedback loops.
6. Algorithmic Collusion
One of the most significant competition concerns is the possibility that algorithms may facilitate coordination.
Traditional cartel:
Human competitors → Communication → Agreement → Coordinated prices
Algorithmic coordination may look different:
Independent algorithms → Data observation → Price prediction → Automated responses → Parallel conduct
The legal difficulty is determining when parallel algorithmic behaviour constitutes unlawful coordination and when it merely reflects rational independent adaptation.
The Eturas litigation is particularly relevant because the CJEU considered whether information transmitted through a common online booking system could support an inference of participation in concerted practices under Article 101 TFEU.
7. Algorithmic Self-Preferencing
A vertically integrated cognitive platform may operate both:
- the infrastructure through which consumers search or transact; and
- competing services supplied through that infrastructure.
It can potentially use its algorithm to favour its own products.
For example:
Search platform → Ranking algorithm → Platform's own service → Consumer
while a rival receives:
Search platform → Ranking algorithm → Lower visibility → Fewer consumers
The competition issue is not simply ownership of the algorithm but whether the dominant undertaking uses control over the algorithmic gateway to exclude rivals.
The European Commission's Google Shopping case is a major illustration.
8. Cognitive Gatekeeping
A cognitive platform may become a gatekeeper when consumers and businesses depend upon its AI or digital infrastructure to reach markets.
Examples include:
- search engines;
- app stores;
- cloud platforms;
- operating systems;
- digital advertising systems;
- AI assistants;
- online marketplaces;
- payment systems.
The EU has adopted an ex ante approach through the DMA. As of the current framework, designated gatekeepers include Alphabet, Amazon, Apple, Booking, ByteDance, Meta and Microsoft across specified core platform services.
9. Competition Governance
Competition governance in cognitive economies should not rely exclusively upon conventional antitrust litigation.
It may involve five interconnected mechanisms:
1. Ex post competition enforcement
Traditional:
- abuse of dominance;
- cartel enforcement;
- merger control;
- exclusionary agreements.
2. Ex ante regulation
Rules imposed before competitive harm becomes irreversible.
The DMA is an important example.
3. Technical governance
Authorities may require:
- interoperability;
- API access;
- data portability;
- auditability;
- transparency;
- non-discrimination.
4. Merger governance
Authorities must examine acquisitions involving:
- AI startups;
- datasets;
- cloud infrastructure;
- foundation models;
- semiconductor technologies;
- AI distribution channels.
5. Institutional monitoring
Continuous monitoring may be necessary because cognitive systems can change their behaviour dynamically.
10. At Least Six Important Case Laws
Case 1 — Google Search (Google Search / United States v Google)
United States v. Google LLC, 2024
The U.S. District Court for the District of Columbia found Google liable under §2 of the Sherman Act for monopolization of relevant general-search and general-search-text-advertising markets and considered Google's distribution agreements to have exclusionary effects.
Relevance to cognitive economies
Search is fundamentally a cognitive service: algorithms determine what information users see.
The case demonstrates the competition significance of:
- defaults;
- distribution agreements;
- data advantages;
- scale;
- search quality;
- network effects; and
- control over access points.
Principle
Control over an algorithmic gateway can create or reinforce durable market power.
Case 2 — Google Shopping
Google and Alphabet v European Commission — Google Shopping, T-612/17
The case concerned Google's preferential treatment of its own comparison-shopping service within general search results.
The General Court upheld the Commission's finding that Google's conduct constituted an abuse of dominant position.
The case is especially relevant because contemporary EU competition analysis expressly recognises network effects as relevant to exclusionary effects in digital markets.
Cognitive-economy significance
The algorithm does not merely process information; it allocates commercial visibility.
Thus:
Algorithmic ranking → Visibility → Traffic → Transactions → Data → Stronger ranking capability
This can create a feedback mechanism favouring the incumbent.
Case 3 — Google Android
Google and Alphabet v European Commission, C-738/22 P
The 2026 CJEU judgment concerns Google's Android-related practices, including contractual restrictions, tying, exclusive pre-installation payments and restrictions affecting Android forks.
Relevance
Android demonstrates how cognitive ecosystems can extend across multiple layers:
- operating system;
- search;
- app stores;
- applications;
- advertising;
- AI services.
A company controlling an operating-system layer may possess significant leverage over downstream cognitive services.
Principle
Competition must be assessed across interconnected technological layers rather than examining each digital product in complete isolation.
Case 4 — Microsoft
Microsoft Corp. v Commission, T-201/04
The Microsoft litigation concerned Microsoft's conduct involving its operating system and interoperability with competing work-group server products.
The case became an important authority concerning:
- interoperability;
- refusal to supply;
- technological tying;
- network effects; and
- leveraging dominance from one technological market into another.
The later Google litigation continues to refer to Microsoft when analysing monopoly power and exclusionary conduct.
Cognitive-economy relevance
Modern AI ecosystems reproduce some of these structural issues:
Core infrastructure → interoperability → downstream applications → competitive access
Therefore, interoperability can become a competition-governance instrument.
Case 5 — Amazon Marketplace / Buy Box
European Commission — Amazon Marketplace and Amazon Buy Box, AT.40462 and AT.40703
The European Commission investigated Amazon's use of non-public marketplace seller data and the operation of the Buy Box.
The Commission ultimately accepted commitments in the proceedings.
Cognitive-economy significance
This case illustrates a distinctive platform problem:
Platform operator + marketplace operator + data processor + competing seller
The platform can potentially learn from independent businesses while simultaneously competing against those businesses.
Principle
Competition governance must examine dual-role data advantages.
Case 6 — Eturas
Eturas and Others v Lietuvos Respublikos konkurencijos taryba, C-74/14
The CJEU considered an online booking system used by travel agencies and the transmission of information through that system concerning restrictions on discounts.
Importance
Eturas is highly relevant to algorithmic coordination because the technological system can become a mechanism through which market participants receive information capable of facilitating coordinated behaviour.
Principle
Technology-mediated communication can be relevant to establishing concerted practices even where coordination does not resemble a conventional face-to-face cartel.
Case 7 — Epic Games v Apple
Epic Games, Inc. v Apple Inc.
Epic challenged Apple's App Store distribution, in-app payment and anti-steering restrictions.
The district court rejected Epic's principal federal antitrust claims but found Apple's anti-steering restrictions unlawful under California's Unfair Competition Law; the Ninth Circuit substantially affirmed the decision.
Cognitive-economy relevance
The App Store illustrates the role of a platform as:
- infrastructure provider;
- marketplace;
- payment intermediary;
- rule-maker; and
- gatekeeper.
AI assistants and cognitive services may increasingly operate through similar platforms.
11. Emerging AI-Specific Competition Issues
The competition issues in cognitive economies are expanding beyond traditional digital platforms.
A. Foundation-model concentration
Competition concerns may arise where a small number of firms control:
- advanced models;
- compute;
- cloud infrastructure;
- specialised chips;
- training datasets;
- AI distribution channels.
The European Commission has identified vertical integration and partnerships involving AI models, cloud infrastructure and downstream services as important competition considerations in generative AI markets.
B. Cloud–AI integration
A cloud provider may simultaneously supply:
Cloud → Compute → Foundation model → AI tools → Enterprise distribution
This creates possible vertical advantages.
The European Commission in June 2026 reached a preliminary position that AWS and Microsoft Azure should be designated as gatekeepers for cloud services, citing factors including entrenched positions, switching costs, ecosystems and the increasing role of AI tools and partnerships in cloud procurement.
12. AI Interoperability
A cognitive economy requires interoperability between:
- AI assistants;
- operating systems;
- applications;
- cloud systems;
- databases;
- search engines;
- digital identity systems.
In July 2026, the European Commission adopted binding measures concerning Google's Android interoperability for competing AI services, seeking to ensure that competing AI providers receive effective access to relevant Android features.
This demonstrates a movement from:
Competition after exclusion
towards:
Technical rules designed to preserve contestability before exclusion becomes entrenched.
13. Data Access as a Competition Remedy
A dominant AI system may possess data unavailable to competitors.
This creates an important distinction between:
Data ownership
Who legally owns or controls the information?
and
Competitive data access
Whether denying access produces competitive foreclosure.
The EU's 2026 Google Search-data proceedings illustrate this issue. The Commission's measures concern access by eligible third-party search engines to anonymised Google Search data under fair, reasonable and non-discriminatory conditions.
14. Algorithmic Discrimination
A cognitive platform can discriminate through:
- ranking;
- recommendation;
- advertising allocation;
- search visibility;
- pricing;
- access conditions;
- eligibility decisions.
The discriminatory effect may be difficult to detect because the decision is produced by a complex model.
Competition governance therefore requires attention to:
- input data;
- model architecture;
- optimisation objectives;
- ranking criteria;
- feedback loops;
- treatment of rivals; and
- actual competitive effects.
15. Algorithmic Pricing
Algorithms can independently adjust prices in response to market information.
Potential scenarios include:
Scenario A — Independent adaptation
Two algorithms independently respond to supply and demand.
This does not automatically establish a cartel.
Scenario B — Common algorithm
Competitors use the same pricing system or intermediary.
The common technological infrastructure may facilitate coordination.
Scenario C — Explicit coordination
Human firms communicate or agree on pricing rules and use algorithms to implement the agreement.
This is much closer to conventional cartel conduct.
Scenario D — Autonomous coordination
Algorithms learn that maintaining parallel pricing maximises returns without explicit human communication.
This presents a difficult frontier for traditional antitrust doctrine.
16. Merger Control in Cognitive Economies
Traditional merger analysis often focuses on:
- market shares;
- concentration;
- unilateral effects;
- coordinated effects.
Cognitive markets require additional examination of:
- datasets;
- AI talent;
- compute access;
- model capabilities;
- APIs;
- cloud dependence;
- distribution channels;
- interoperability;
- potential competition.
A small AI company may have relatively low current revenue but possess strategically important:
technology + data + talent + intellectual property + future competitive potential.
Consequently, transaction-value and qualitative theories of harm can become important.
17. Killer Acquisitions and Nascent AI Competitors
A dominant platform might acquire an emerging AI company before it becomes a significant competitor.
Competition authorities may therefore ask:
- Was the target developing a competing technology?
- Could it have become an independent competitor?
- Does the acquisition eliminate a future source of innovation?
- Does the transaction consolidate access to critical data?
- Does it strengthen the acquirer's ecosystem?
- Does it increase barriers to entry?
The question is increasingly about innovation competition, not merely current market share.
18. Consumer Autonomy and Competition
Cognitive economies also blur the distinction between competition law and consumer protection.
AI systems may:
- personalise prices;
- predict willingness to pay;
- optimise advertisements;
- influence purchasing decisions;
- determine search rankings;
- recommend products.
Therefore, competitive analysis may need to consider whether consumers can realistically:
- switch providers;
- understand ranking systems;
- access alternatives;
- control their data;
- change defaults;
- use competing AI systems.
Recent EU scholarship specifically identifies behavioural economics as relevant to both Article 102 TFEU enforcement and the DMA in digital markets.
19. Competition Remedies
Traditional remedies include:
- fines;
- injunctions;
- divestiture;
- prohibition of agreements.
Cognitive economies may require more technically sophisticated remedies.
A. Interoperability
Require competing systems to communicate.
B. Data portability
Allow users or businesses to transfer relevant data.
C. API access
Prevent discriminatory denial of technical access.
D. Non-discrimination
Require neutral treatment in ranking and access.
E. Data separation
Prevent competitively sensitive information from being improperly transferred between platform functions.
F. Algorithmic auditing
Permit regulators or independent experts to examine relevant system behaviour.
G. Structural separation
In extreme cases, separate infrastructure from downstream competitive activities.
20. Competition Governance Model for Cognitive Economies
A useful framework is:
Layer 1 — Market structure
Examine:
- concentration;
- entry barriers;
- network effects;
- economies of scale.
↓
Layer 2 — Cognitive resources
Examine:
- data;
- models;
- compute;
- algorithms;
- talent.
↓
Layer 3 — Platform control
Examine:
- defaults;
- rankings;
- APIs;
- operating systems;
- app stores.
↓
Layer 4 — Behaviour
Examine:
- exclusion;
- tying;
- self-preferencing;
- discrimination;
- refusal to deal;
- algorithmic coordination.
↓
Layer 5 — Competitive effects
Examine:
- foreclosure;
- innovation;
- quality;
- consumer choice;
- entry;
- interoperability.
↓
Layer 6 — Governance remedy
Apply:
- antitrust enforcement;
- merger control;
- interoperability;
- data access;
- transparency;
- ex ante regulation.
21. Relationship Between Competition Law and AI Governance
Competition law should not attempt to regulate every AI risk.
There should be a distinction between:
| Issue | Primary concern |
|---|---|
| Algorithmic cartel | Competition law |
| AI safety | AI/safety regulation |
| Data privacy | Data-protection law |
| Discriminatory AI | Equality/consumer law |
| Dominant AI platform | Competition law |
| Foundation-model concentration | Competition + sectoral regulation |
| AI interoperability | Competition/digital regulation |
| AI merger | Merger control |
| AI-generated misinformation | Other regulatory frameworks |
Nevertheless, these regimes increasingly overlap.
22. Major Legal Challenges
1. Defining the relevant market
AI systems may compete across traditional market boundaries.
For example, an AI assistant can potentially substitute for:
- search;
- software;
- shopping;
- customer service;
- advertising;
- information services.
Traditional market-definition methodologies may therefore become difficult to apply.
2. Measuring market power
Market power may depend upon:
- data;
- compute;
- model quality;
- ecosystem position;
- switching costs.
Revenue and market share alone may not capture the full competitive position.
3. Proving algorithmic causation
A regulator must distinguish:
algorithmic design → conduct → competitive harm
from:
algorithmic output → coincidence → competitive harm
This creates substantial evidentiary difficulties.
4. Explainability
Competition authorities may need technical evidence concerning:
- training data;
- model objectives;
- ranking variables;
- recommendation systems;
- optimisation functions;
- logs.
5. Speed of technological change
A competition investigation may take years while AI technology can change within months.
This strengthens the case for appropriate interim and ex ante governance mechanisms.
23. Six Core Doctrinal Principles
The emerging law of cognitive economies can be organised around six principles:
Principle 1 — Data is potentially a strategic competitive asset
But possession alone does not establish an infringement.
Principle 2 — Algorithmic control can constitute market power
Particularly where the algorithm controls access to consumers or commercial opportunities.
Principle 3 — Interoperability can preserve contestability
Technological restrictions may prevent rivals from effectively competing.
Principle 4 — Network effects can reinforce dominance
Particularly where increased use generates additional data that improves the system.
Principle 5 — Cognitive systems can facilitate coordination
Common algorithms and technological infrastructures may alter the analysis of concerted practices.
Principle 6 — Competition governance must protect innovation
The objective is not simply lower prices but preservation of:
- innovation;
- choice;
- access;
- entry;
- quality;
- technological diversity.
24. Comparative Case-Law Matrix
| Case | Jurisdiction | Principal issue | Cognitive-economy lesson |
|---|---|---|---|
| Google Search | USA | Search monopolization | Algorithmic distribution and defaults |
| Google Shopping | EU | Self-preferencing | Algorithmic ranking can affect rivals |
| Google Android | EU | Tying/exclusion | Ecosystem leverage and interoperability |
| Microsoft | EU | Interoperability | Control of technological infrastructure |
| Amazon Marketplace/Buy Box | EU | Platform/data practices | Dual-role platform and data advantage |
| Eturas | EU/Lithuania | Technology-assisted coordination | Digital systems can facilitate concerted practices |
| Epic Games v Apple | USA | App Store restrictions | Platform gatekeeping and anti-steering |
| United States v Google | USA | Search and advertising monopolization | Distribution agreements can reinforce cognitive-network power |
25. Conclusion
Competition law in cognitive economies represents a transition from regulating predominantly physical and price-based markets to regulating information-intensive, algorithmically mediated and continuously learning markets.
The central competitive resources increasingly include data, algorithms, compute, AI models, interoperability and access to digital ecosystems.
The case law of Microsoft, Google Shopping, Google Android, Google Search, Amazon Marketplace, Eturas and Epic Games v Apple demonstrates that many of the fundamental competition-law principles remain applicable. However, their application increasingly requires attention to technological architecture, data advantages, network effects, algorithmic decision-making and platform governance.
The emerging model can therefore be expressed as:
Cognitive competition governance = antitrust enforcement + merger control + interoperability + data governance + algorithmic oversight + ex ante platform regulation.
The development of the EU DMA is particularly significant because it supplements traditional ex post competition enforcement with obligations designed to preserve contestability and fairness in digital ecosystems. Current EU measures concerning AI interoperability, search-data access and cloud services demonstrate that competition governance is increasingly moving into the technical architecture of cognitive markets itself.

comments