Competition Law and Competition Governance in Intelligence Economies

1. Introduction

An intelligence economy is an economy in which competitive advantage increasingly depends upon the ability to collect, process, predict and act upon information through artificial intelligence (AI), machine learning, algorithms, data analytics, cloud computing, foundation models, automated decision-making and intelligent agents.

Unlike a traditional digital economy, where data and platforms are often the principal competitive assets, an intelligence economy adds a further layer: computational intelligence. Firms compete not merely through possession of data but through the ability to convert data into predictions, recommendations, automated decisions and autonomous actions.

This creates new competition-law questions:

• Who controls the data necessary to train intelligent systems?
• Who controls computing capacity and cloud infrastructure?
• Can dominant firms use AI to extend their dominance into adjacent markets?
• Can competing algorithms coordinate prices without an explicit human agreement?
• Can AI-driven ranking systems discriminate against rivals?
• Can foundation-model providers restrict interoperability?
• Can cloud providers tie computing services to their own AI models?
• How should merger control address acquisitions of AI startups and talent?
• How should competition authorities investigate systems whose decision-making processes are difficult to explain?

The OECD has identified both the pro-competitive potential of AI—such as lower entry barriers and productivity gains—and risks involving data access, model restrictions, algorithmic collusion and reduced contestability.

2. Meaning and Characteristics of an Intelligence Economy

An intelligence economy can be understood as a market environment where intelligence itself becomes an important economic input and competitive asset.

Its principal characteristics include:

A. Data-driven competition

Data can function as:

• an input into AI training;
• a source of consumer insight;
• a means of improving prediction;
• a mechanism for personalisation;
• a barrier to entry; and
• a source of feedback effects.

A firm with a large user base can generate more data, improve its model, attract more users and consequently generate still more data.

This produces a potential data-feedback loop:

Users → Data → Training → Better Model → Better Service → More Users → More Data.

Competition authorities therefore increasingly have to examine whether control over data creates durable market power.

B. Computing power as a competitive input

AI development depends upon computing resources, including:

• GPUs;
• specialised AI accelerators;
• cloud infrastructure;
• data centres;
• storage;
• networking;
• model-training infrastructure.

Consequently, competition may arise not only at the AI-model level but also at the compute layer.

A vertically integrated firm controlling cloud infrastructure and an AI model may have incentives to:

• prioritise its own models;
• offer preferential computing access;
• impose technical restrictions;
• create switching costs;
• bundle cloud and AI products; or
• enter exclusive arrangements.

The European Commission's 2026 preliminary assessment concerning AWS and Azure illustrates the growing competition significance of cloud infrastructure and AI ecosystems, particularly because AI tools and partnerships can affect cloud procurement and ecosystem lock-in.

3. Competition Governance

Competition governance in an intelligence economy means the institutional framework through which governments, competition authorities, regulators and courts ensure that intelligent technologies develop within competitive markets.

It involves five principal dimensions:

1. Ex ante regulation
2. Antitrust enforcement
3. Merger control
4. Market investigation and monitoring
5. Technological and algorithmic governance

The objective is not to prevent AI development. Rather, competition governance attempts to ensure that technological development does not become a mechanism for permanently excluding competitors.

4. Major Competition Concerns

4.1 Algorithmic Collusion

One of the most important issues is whether algorithms can facilitate coordination between competitors.

Traditional cartel law normally looks for:

• an agreement;
• communication;
• coordination;
• exchange of commercially sensitive information; or
• concerted practices.

Algorithms complicate this analysis.

Competitors may use the same pricing software or a common algorithmic provider. The system may then repeatedly adjust prices in response to competitors.

The OECD has specifically identified the risk that common pricing algorithms and hub-and-spoke arrangements can facilitate coordination, including situations in which competitors use identical repricing software.

Competition-law questions

Authorities may need to determine:

• Who designed the algorithm?
• What instructions were given to it?
• What information does it receive?
• Does it access competitors' sensitive information?
• Does it autonomously monitor competitors?
• Was there human communication?
• Could the undertaking reasonably anticipate coordinated effects?

The critical distinction is between independent algorithmic adaptation and algorithmic implementation of an underlying anti-competitive agreement.

5. Algorithmic Discrimination and Personalised Pricing

Intelligence systems can analyse:

• location;
• purchasing history;
• browsing behaviour;
• income proxies;
• device characteristics;
• search behaviour;
• willingness to pay.

This may allow firms to charge different consumers different prices.

Personalisation is not automatically unlawful. However, competition concerns may arise where a dominant undertaking uses algorithmic discrimination to:

• exclude competitors;
• disadvantage particular classes of users;
• exploit locked-in customers;
• discriminate against business customers;
• prevent market entry.

The regulatory challenge is therefore to distinguish legitimate price differentiation from discriminatory conduct that produces exclusionary or exploitative effects.

6. AI Self-Preferencing

An intelligent platform may control both:

1. the marketplace or platform; and
2. competing products or services offered through that platform.

Its algorithm may then systematically rank its own product more favourably.

Examples could include:

• an AI search engine prioritising its affiliated service;
• an AI shopping assistant recommending affiliated products;
• an app store AI ranking affiliated applications;
• an AI travel platform favouring its own booking service.

This raises questions similar to those encountered in traditional platform self-preferencing cases, but the algorithmic environment can make the conduct less visible.

7. Data Access and Data Monopolisation

A dominant intelligent system may benefit from exclusive access to large datasets.

Competition authorities may therefore examine:

Input foreclosure

A dominant firm prevents rivals from obtaining essential data.

Data degradation

Data is technically made available but at inferior quality.

Discriminatory access

Affiliated businesses receive superior access.

Exclusive data arrangements

Contracts prevent customers or suppliers from providing data to competitors.

Data portability restrictions

Users cannot easily transfer relevant information to competing services.

The question is not simply whether data is valuable. The important issue is whether control over particular data creates substantial and durable competitive advantages that rivals cannot reasonably reproduce.

8. Foundation Models and Competition

Foundation models represent a particularly important layer of intelligence economies.

They can be used across:

• search;
• healthcare;
• finance;
• education;
• software;
• logistics;
• robotics;
• advertising;
• customer service.

The CMA's foundation-model review specifically examined how these markets could evolve and identified risks to fair, open and effective competition.

Three structural bottlenecks are particularly important:

Data → Compute → Distribution

A firm controlling all three can potentially obtain significant ecosystem advantages.

9. Cloud–AI Vertical Integration

Cloud providers may simultaneously operate:

• cloud infrastructure;
• AI chips;
• foundation models;
• developer tools;
• enterprise software;
• application marketplaces.

This creates potential vertical competition concerns.

Possible theories of harm include:

• tying;
• bundling;
• discriminatory access;
• exclusive contracts;
• interoperability restrictions;
• preferential cloud access;
• foreclosure of competing AI developers.

The competition analysis must distinguish legitimate integration efficiencies from conduct that prevents rival AI providers from obtaining viable access to infrastructure.

10. AI Mergers and Acquisitions

Traditional merger thresholds may be inadequate where the target AI company:

• has little current revenue;
• possesses important technology;
• has highly skilled personnel;
• owns strategically valuable datasets;
• controls an emerging technology;
• represents a potential future competitor.

Consequently, competition authorities increasingly examine transactions involving:

• AI startups;
• foundation-model companies;
• cloud providers;
• semiconductor businesses;
• AI-chip companies;
• data providers.

Recent CMA investigations have included transactions and partnerships involving Microsoft/Inflection, Microsoft/Mistral AI, Amazon/Anthropic and Alphabet/Anthropic.

11. Six Important Case Laws and Enforcement Examples

1. United States v. Google LLC — Search and Search Distribution

Jurisdiction: United States
Court: U.S. District Court for the District of Columbia

The Google search litigation concerns alleged exclusionary practices used to maintain Google's position in general search and search advertising.

Relevance to intelligence economies

Search is increasingly an AI-enabled information service. Search data, distribution and user interaction can provide important inputs for developing intelligent systems.

The case therefore illustrates the relationship between:

Distribution → Data → Search quality → AI capability → Market power

The later remedies debate has also recognised the competitive significance of search data for AI competitors. The OECD reported in 2025 that proposed remedies in the Google Search case included data-sharing measures relevant to AI competition.

Principle

Control over a major information gateway can have competitive consequences extending beyond the immediate market.

2. FTC v. Amazon

Jurisdiction: United States
Authority: Federal Trade Commission and State Attorneys General

The FTC alleges that Amazon employed interconnected strategies that unlawfully maintained monopoly power in online retail and related markets.

Intelligence-economy relevance

Amazon's marketplace demonstrates how algorithms can control:

• product visibility;
• pricing;
• seller access;
• consumer recommendations;
• marketplace participation.

An intelligent marketplace can therefore become both:

marketplace + algorithmic gatekeeper.

Principle

Competition analysis increasingly needs to examine the architecture of algorithmically controlled platforms rather than only their headline prices.

3. United States v. Apple Inc.

Jurisdiction: United States
Authority: U.S. Department of Justice

The case concerns allegations concerning Apple's practices relating to the iPhone ecosystem.

Intelligence-economy relevance

The case illustrates ecosystem-based competition involving:

• operating systems;
• application distribution;
• APIs;
• interoperability;
• hardware-software integration;
• developer access.

These issues become particularly significant when intelligent assistants and AI applications depend upon access to device-level functionality.

Principle

Control over an ecosystem can provide leverage over adjacent markets and emerging technologies.

4. European Commission — Google Shopping

Jurisdiction: European Union
Legal framework: Article 102 TFEU

The Google Shopping matter concerned Google's preferential treatment of its own comparison-shopping service within general search results.

Intelligence-economy relevance

The underlying concept is highly relevant to AI-driven search and recommendation systems.

An AI assistant may increasingly determine:

• which products users see;
• which services are recommended;
• which suppliers receive visibility.

Thus, the traditional search self-preferencing problem can evolve into AI recommendation self-preferencing.

Principle

A dominant information intermediary may face competition-law scrutiny where it systematically advantages its own downstream service.

5. Trod Ltd / GB Eye Ltd

Jurisdiction: United Kingdom
Authority: Competition and Markets Authority

The CMA's Trod/GB Eye case is an important algorithmic-cartel example.

The businesses used automated repricing software in an online marketplace environment.

Intelligence-economy relevance

It demonstrates that algorithms can become instruments through which an existing anti-competitive arrangement is implemented.

The OECD identifies the case as an important example in the development of competition-law thinking about algorithmic pricing.

Principle

The use of an algorithm does not immunise conduct from ordinary competition law.

6. Eturas v Lietuvos Respublikos Konkurencijos Taryba

Court: Court of Justice of the European Union
Case: C-74/14

The case concerned an electronic booking platform and an electronically communicated restriction affecting discounts.

Intelligence-economy relevance

The importance of the case lies in demonstrating that digital systems can facilitate coordination among multiple businesses.

The legal question is not necessarily whether managers physically meet in a traditional setting, but whether businesses knowingly participate in a mechanism that facilitates coordinated conduct.

Principle

Competition law can apply to technologically mediated coordination just as it applies to traditional forms of communication.

12. Additional Important Authorities

For an advanced intelligence-economy research framework, the following authorities are also relevant:

Google Android

Relevant to:

• tying;
• mobile ecosystems;
• default settings;
• application distribution;
• leveraging dominance.

Google AdSense

Relevant to:

• advertising intermediation;
• exclusivity;
• platform power;
• data advantages.

Meta / Facebook

Relevant to:

• data advantages;
• platform ecosystems;
• acquisitions;
• privacy and competition interactions.

Microsoft / Activision Blizzard

Relevant to:

• ecosystem expansion;
• cloud gaming;
• vertical foreclosure;
• access to important content.

The CMA's investigation demonstrates how modern merger control can involve cloud infrastructure and ecosystem effects rather than simply traditional horizontal overlaps.

13. Competition Governance Model for Intelligence Economies

A comprehensive governance framework can be represented as:

DATA

↓

COMPUTE

↓

FOUNDATION MODELS

↓

AI APPLICATIONS

↓

DIGITAL PLATFORMS

↓

CONSUMERS / BUSINESSES

At every level, competition authorities should examine:

Access → Interoperability → Pricing → Ranking → Data → Switching → Entry → Innovation

14. Role of Competition Authorities

Competition authorities should develop specialised capabilities for:

A. Algorithmic auditing

Authorities need technical capacity to determine how algorithms operate.

B. Data analysis

Authorities increasingly need access to:

• datasets;
• model outputs;
• training information;
• logs;
• pricing histories;
• ranking systems.

C. Merger monitoring

Authorities should examine potential competition, not merely current market shares.

D. Interoperability assessment

Technical interoperability can determine whether rivals can realistically compete.

E. Market investigations

Where traditional enforcement is too slow, sector-wide market studies may identify structural risks.

The CMA's AI Foundation Models programme demonstrates this preventive approach: rather than waiting for a completed infringement, it examined the emerging structure of foundation-model markets and formulated principles concerning competition and consumer protection.

15. Ex Ante and Ex Post Regulation

Ex Post

Traditional competition law intervenes after potentially harmful conduct occurs.

Examples:

• abuse of dominance;
• cartels;
• exclusionary agreements;
• anti-competitive mergers.

Ex Ante

Intelligence economies increasingly justify preventive mechanisms such as:

• interoperability obligations;
• data-access requirements;
• transparency requirements;
• switching rights;
• restrictions on self-preferencing;
• gatekeeper obligations;
• merger notification mechanisms.

The EU Digital Markets Act represents an important example of an ex-ante approach to certain digital gatekeepers.

16. Competition and Innovation

AI creates a difficult regulatory balance.

Excessive intervention could potentially:

• discourage investment;
• reduce incentives to develop new models;
• increase compliance costs;
• slow experimentation.

Insufficient intervention could potentially allow:

• entrenched incumbency;
• concentration of compute;
• control over critical datasets;
• exclusion of startups;
• ecosystem lock-in;
• algorithmic coordination.

Accordingly, competition governance should preserve contestability while allowing technological experimentation.

The CMA itself describes foundation models as general-purpose technology capable of affecting many sectors, while emphasising the importance of competitive markets for innovation and economic benefits.

17. Key Doctrines Applicable to Intelligence Economies

18. Emerging Issues

18.1 Autonomous AI Agents

Future AI agents may independently:

• negotiate prices;
• purchase goods;
• select suppliers;
• enter contracts;
• change prices;
• allocate resources.

Competition law may therefore have to determine who is responsible for autonomous competitive conduct.

18.2 Algorithmic Tacit Coordination

Self-learning algorithms could potentially converge on similar market outcomes without explicit communication.

The OECD has specifically identified algorithmic collusion and self-learning pricing systems as emerging competition concerns.

This creates a difficult question:

When does rational algorithmic adaptation become legally problematic coordination?

18.3 AI and Essential Facilities

Potentially strategic infrastructure may include:

• specialised GPUs;
• cloud computing;
• datasets;
• model interfaces;
• AI marketplaces;
• foundation models.

Competition law may therefore need to revisit the relationship between essential-facility principles and intangible technological infrastructure.

18.4 AI and Consumer Choice

AI intermediaries may increasingly make choices for consumers.

Instead of consumers comparing ten products, an AI assistant may recommend one.

This creates a new competition problem:

Who controls the recommendation layer?

If one AI intermediary becomes the principal decision-maker between consumers and suppliers, control over recommendations may become an important source of market power.

19. Compliance Framework for Businesses

Businesses operating in intelligence economies should establish an AI Competition Compliance Programme.

It should include:

1. Algorithm governance

Document the objectives and parameters of pricing and recommendation algorithms.

2. Competition audit

Regularly test algorithms for exclusionary or discriminatory outcomes.

3. Information controls

Prevent competitors' commercially sensitive information from being improperly incorporated into shared systems.

4. AI procurement review

Review third-party algorithmic and pricing software.

5. Merger review

Assess competition implications of AI partnerships and acquisitions.

6. Data governance

Ensure data-sharing arrangements do not create unlawful information exchange.

7. Human oversight

Maintain appropriate human review of high-impact competitive decisions.

8. Documentation

Preserve:

• model versions;
• instructions;
• datasets;
• system logs;
• pricing decisions;
• ranking changes;
• communications with vendors.

This evidence can become critical during a competition investigation.

20. Future of Competition Governance

The next stage of competition governance is likely to move from firm-centred regulation to ecosystem-centred regulation.

The relevant competitive unit may increasingly be:

Data + Compute + Model + Platform + Distribution

rather than simply an individual company.

Competition authorities will therefore need interdisciplinary capabilities involving:

• competition law;
• economics;
• computer science;
• data science;
• AI engineering;
• cybersecurity;
• consumer protection;
• intellectual property;
• privacy law.

21. Conclusion

Competition law in intelligence economies extends traditional antitrust principles into an environment where data, algorithms, computing power, AI models and autonomous decision-making become central competitive assets.

The principal risks include:

1. algorithmic collusion;
2. data foreclosure;
3. AI self-preferencing;
4. cloud–AI leveraging;
5. interoperability restrictions;
6. algorithmic discrimination;
7. ecosystem lock-in;
8. anti-competitive AI mergers; and
9. control over foundation-model infrastructure.

The existing cases involving Google, Amazon, Apple, Trod/GB Eye and Eturas demonstrate that established competition-law doctrines can already address many technologically mediated forms of conduct. At the same time, newer AI foundation-model and cloud investigations show that competition governance is moving toward continuous market monitoring and preventive intervention.