Competition Law and Future Intelligence Monopolization Risks

Introduction

Intelligence monopolization refers to the possibility that a small number of firms could obtain durable control over critical artificial-intelligence capabilities, infrastructure, data, computing resources, models, interfaces, or autonomous commercial networks. Unlike traditional monopolies, future intelligence markets may be characterized by ecosystem control rather than simple ownership of a single product.

Competition law therefore faces a broader question: how should antitrust law respond when economic power is accumulated through control of AI models, compute, proprietary data, algorithms, AI agents, distribution channels, cloud infrastructure, semiconductor supply, or interoperable ecosystems?

Traditional competition law remains applicable through concepts such as:

• dominance and monopolization;
• abuse of market power;
• exclusionary conduct;
• tying and bundling;
• refusal of access;
• discriminatory access;
• exclusive dealing;
• predatory pricing;
• merger control;
• essential facilities;
• interoperability;
• data-related foreclosure;
• vertical restraints; and
• coordinated conduct.

However, AI creates additional difficulties because scale, learning effects, network effects, switching costs, data advantages and computational resources can reinforce one another.

1. Meaning of Intelligence Monopolization

Intelligence monopolization can occur where a firm acquires or maintains substantial control over one or more strategic layers of the AI economy.

A. Compute monopolization

A firm may control access to:

• advanced AI accelerators;
• data centres;
• cloud computing;
• specialized chips;
• high-bandwidth networking;
• model-training infrastructure.

If competitors cannot obtain comparable computing resources, the compute layer may become a bottleneck.

B. Data monopolization

AI systems may depend upon:

• proprietary datasets;
• behavioural data;
• transaction data;
• industrial data;
• search data;
• location data;
• scientific datasets;
• user-generated content.

Exclusive accumulation of strategically important data can create barriers to entry.

C. Model monopolization

A highly capable foundation model may become a central platform through which numerous downstream services operate.

The risk increases where the model provider also controls:

• applications;
• cloud services;
• operating systems;
• app stores;
• search;
• advertising;
• enterprise software.

D. Agentic monopolization

Future autonomous AI agents could execute:

• purchases;
• financial transactions;
• procurement;
• logistics;
• advertising;
• negotiations;
• employment decisions;
• software development.

Control over the agent layer could therefore provide influence over several downstream markets simultaneously.

2. Why AI Creates Special Competition Risks

2.1 Network Effects

AI platforms may become more valuable as more users, developers and businesses join them.

More users generate more interaction data.

More data may improve the system.

Better performance attracts more users.

This creates a potentially self-reinforcing cycle:

Users → Data → Better AI → More Users → Greater Data → Greater AI Advantage

Such feedback mechanisms can make markets difficult for new entrants to penetrate.

3. Economies of Scale

Training frontier AI models can require enormous expenditures on:

• computing;
• electricity;
• engineering;
• data acquisition;
• research;
• specialised hardware;
• safety testing.

Consequently, average costs may decline substantially as the scale of operation increases.

This may create a structural advantage for incumbent firms.

Competition authorities therefore need to distinguish between:

legitimate economies of scale

and

strategic exclusion designed to prevent competitors from achieving scale.

4. Data as a Competitive Asset

Traditional competition analysis often focuses on price, output and market shares.

AI requires greater attention to quality, data access and innovation.

A dominant platform might obtain data unavailable to rivals because it controls:

• search;
• social media;
• e-commerce;
• payments;
• cloud computing;
• mobile operating systems;
• advertising;
• productivity software.

The resulting advantage can become cumulative.

Possible competition concern

A dominant firm may use data collected in one market to strengthen its position in another market.

This can create cross-market leveraging.

5. Vertical Integration and Intelligence Monopolization

AI markets can involve multiple vertically connected layers:

Semiconductors → Cloud → Compute → Foundation Model → AI Platform → Applications → Distribution

A firm controlling several layers may have incentives to disadvantage competitors.

Examples include:

• preferential cloud access for its own model;
• discriminatory pricing of compute;
• technical restrictions on competing models;
• tying AI services to operating systems;
• exclusive agreements with developers;
• preferential placement of proprietary AI products.

Competition law must therefore examine ecosystem foreclosure, not merely individual transactions.

6. Tying and Bundling

A powerful AI provider could condition access to one product upon purchase or use of another.

Examples:

• AI assistant + operating system;
• AI model + cloud services;
• AI search + advertising;
• AI enterprise software + database services;
• AI coding assistant + development platform.

The traditional tying framework remains relevant, but AI may make the analysis more complicated because products can be technologically integrated rather than contractually separate.

7. Exclusive Dealing and AI Ecosystems

AI firms may enter agreements requiring customers or developers to use their:

• cloud platform;
• foundation model;
• API;
• data infrastructure;
• AI accelerator;
• deployment environment.

Exclusive arrangements may produce efficiencies, including investment incentives.

However, they can also foreclose competing AI providers if used extensively by a dominant undertaking.

8. Interoperability and Switching Costs

AI ecosystems may become difficult to leave because users accumulate:

• prompts;
• workflows;
• fine-tuned models;
• proprietary data;
• agent configurations;
• application integrations;
• organizational knowledge.

This creates AI switching costs.

A competition authority may therefore examine whether a dominant provider artificially prevents:

• data portability;
• model portability;
• API interoperability;
• migration;
• compatibility with competing AI systems.

9. Essential Facilities and AI

Some AI resources may become sufficiently important to downstream competition that competitors argue for access.

Potential bottlenecks could include:

• specialized compute;
• critical datasets;
• cloud infrastructure;
• model interfaces;
• technical standards;
• digital identity systems;
• AI distribution channels.

The essential-facilities doctrine must nevertheless be applied cautiously because compulsory access can reduce incentives to invest.

The central question is whether the resource is genuinely indispensable and whether denial of access produces competitive harm rather than merely making competition more difficult.

10. Merger Control and Intelligence Monopolization

Traditional merger thresholds may fail to capture acquisitions involving AI startups.

A startup may have:

• low revenue;
• few employees;
• valuable technology;
• proprietary datasets;
• important researchers;
• strategic intellectual property.

Consequently, an acquisition may be competitively significant even before the target generates substantial turnover.

Competition authorities may therefore examine:

• innovation competition;
• nascent competitors;
• data assets;
• intellectual property;
• AI researchers;
• potential competition;
• access to computing resources;
• vertical foreclosure.

11. Killer Acquisitions in AI

A dominant AI company may acquire an emerging firm whose technology could eventually become a competitive threat.

The concern is not necessarily the target's current market share.

Instead, the question may be whether the target represents:

an important future source of competitive innovation.

This makes traditional static market-share analysis insufficient by itself.

12. AI and Predatory Pricing

AI services may initially be offered:

• free;
• below cost;
• heavily subsidized;
• bundled with unrelated services.

A powerful diversified company can potentially absorb losses in AI while weakening independent competitors.

Competition authorities must distinguish:

pro-competitive introductory pricing

from

strategic exclusionary pricing.

The analysis should consider:

• duration;
• recoupment possibilities;
• market structure;
• internal documents;
• switching effects;
• cross-subsidization;
• foreclosure of rivals.

13. Algorithmic Self-Preferencing

An AI platform may control both:

1. the infrastructure used by competitors; and
2. competing AI applications.

It could potentially rank or recommend its own AI services more favourably.

Examples include:

• preferential search placement;
• preferred API routing;
• better access to system resources;
• default AI assistant placement;
• privileged access to user data.

Self-preferencing therefore becomes particularly important where the platform acts simultaneously as infrastructure provider and competitor.

14. Algorithmic Collusion

AI agents could independently observe competitors' prices and rapidly adjust their own strategies.

This creates a difficult distinction between:

• lawful independent adaptation; and
• unlawful coordination.

Traditional cartel law generally requires evidence of an agreement or concerted practice.

The emergence of autonomous pricing systems raises questions about whether firms can be responsible where algorithms facilitate coordination without conventional human communication.

15. Autonomous AI Agents and Competition

Future AI agents may negotiate with other agents.

For example:

Agent A → negotiates price → Agent B → adjusts offer → Agent C → changes supply

If many firms deploy similar optimization systems, algorithmic interaction could produce stable supracompetitive outcomes.

Competition authorities may therefore need to examine:

• algorithm design;
• training objectives;
• communications between agents;
• pricing parameters;
• information exchange;
• human supervision;
• audit trails.

16. Six Important Case Laws

The following cases do not all concern modern generative AI directly. Their importance lies in the competition-law principles that can be applied to future intelligence markets.

Case 1: United States v. Microsoft Corp. (2001)

The Microsoft litigation concerned Microsoft's dominance in operating systems and its conduct toward competing technologies, particularly web browsers.

The case is significant for intelligence monopolization because it demonstrates how a dominant platform can use control over one technological layer to disadvantage emerging competitors.

Relevance to AI

The Microsoft principles can inform analysis of:

• AI + operating-system integration;
• default AI assistants;
• platform foreclosure;
• tying;
• technical restrictions;
• distribution advantages.

Principle: Control of an important technological platform can provide opportunities for exclusionary conduct in adjacent markets.

Case 2: United States v. Google LLC — Search (2024)

The U.S. federal litigation concerning Google's search distribution practices examined agreements that allegedly reinforced Google's position in general search.

The case illustrates the importance of distribution and default arrangements in maintaining digital-market power.

Relevance to AI

AI providers may similarly seek:

• default status;
• preferred placement;
• exclusive distribution;
• browser integration;
• device integration.

Thus, intelligence competition may depend not merely upon the quality of an AI model but also upon how users reach it.

Case 3: United States v. Google LLC — Ad Tech

The U.S. government has also pursued litigation concerning Google's advertising-technology ecosystem.

The broader significance is the examination of market power across interconnected technological layers.

Relevance to AI

AI ecosystems may similarly involve multiple vertically related markets:

data → model → advertising → distribution → applications.

Competition analysis may therefore need to examine how power at one layer affects competition elsewhere.

Case 4: European Commission v. Google Shopping

The European Commission's Google Shopping decision concerned preferential treatment of Google's comparison-shopping service within its general search results.

The case is particularly relevant to self-preferencing.

Relevance to AI

A dominant AI platform could potentially privilege its own:

• shopping agent;
• travel agent;
• financial assistant;
• coding service;
• marketplace;
• search product.

The Shopping decision demonstrates why discriminatory ranking can become a competition issue where a dominant platform also operates downstream services.

Case 5: Google Android — European Commission

The European Commission's Android decision examined practices involving Google's Android ecosystem, including tying and distribution arrangements.

Relevance to AI

The case provides useful principles for analyzing:

• AI assistants bundled with operating systems;
• default arrangements;
• application distribution;
• tying;
• ecosystem leverage.

An AI provider with control over a major operating system could potentially use that position to strengthen an AI product.

Case 6: European Commission v. Microsoft (Microsoft Tying)

The EU Microsoft proceedings addressed Microsoft's conduct involving interoperability and tying in software markets.

Relevance to AI

Interoperability may become central to future AI competition.

Competition authorities may examine whether a dominant AI ecosystem restricts:

• competing models;
• third-party agents;
• APIs;
• interoperability;
• data portability.

The case therefore provides an important foundation for thinking about technical foreclosure.

17. Additional Important Authorities

Several additional cases are particularly useful for developing the doctrine.

Bronner v Mediaprint

The European Court of Justice established important limitations on compulsory-access theories.

Future relevance: AI infrastructure providers should not automatically be required to provide access merely because competitors depend upon their services.

IMS Health

The case examined refusal to license intellectual property and the exceptional circumstances in which compulsory licensing may be justified.

Future relevance: proprietary AI models, training technologies and datasets could generate similar tensions between intellectual-property rights and competition.

Magill

The Court developed principles concerning exceptional circumstances involving refusal to license intellectual property.

Future relevance: particularly relevant where control of proprietary AI technology becomes indispensable to downstream competition.

Intel v Commission

The case concerned exclusionary rebates and the assessment of competitive effects.

Future relevance: AI/cloud providers offering rebates or incentives to customers could face similar scrutiny where those arrangements foreclose rivals.

Qualcomm

European Commission litigation concerning Qualcomm's practices illustrates the importance of exclusionary payments and market foreclosure in technology markets.

Future relevance: AI chip, cloud and model providers may use financial incentives to secure exclusivity or preferential distribution.

18. Future Forms of Intelligence Monopolization

19. Competition-Law Tests for Future AI Markets

A future competition authority should consider at least six dimensions.

1. Market power

What constitutes the relevant market?

Possible markets include:

• foundation models;
• AI inference;
• AI agents;
• AI cloud services;
• AI chips;
• AI data;
• AI applications.

2. Contestability

Can new firms realistically enter?

The authority should consider:

• capital requirements;
• computing access;
• data;
• talent;
• intellectual property;
• distribution.

3. Ecosystem effects

Does the firm control several adjacent markets?

4. Innovation effects

Does the conduct reduce:

• technological innovation;
• model development;
• research competition;
• product variety?

5. Interoperability

Can users and developers switch to alternatives?

6. Long-term foreclosure

Could apparently small conduct create durable structural dependence?

20. Remedies for Intelligence Monopolization

Competition authorities could potentially employ several remedies.

Structural remedies

In exceptional circumstances:

• divestiture;
• separation of business units;
• restrictions on vertical integration.

Behavioural remedies

Possible measures include:

• non-discrimination obligations;
• interoperability;
• data portability;
• API access;
• prohibition of exclusivity;
• transparent ranking;
• restrictions on tying.

Merger remedies

Authorities may impose:

• divestiture commitments;
• licensing obligations;
• access commitments;
• firewalls;
• restrictions on data combination.

Transparency remedies

AI platforms could be required to maintain records concerning:

• ranking;
• model access;
• pricing;
• algorithmic changes;
• discriminatory treatment.

21. Institutional Challenges for Competition Authorities

Future competition authorities may require capabilities beyond traditional antitrust economics.

They may need expertise in:

• AI engineering;
• machine learning;
• cloud architecture;
• semiconductor technology;
• data governance;
• cybersecurity;
• algorithmic auditing;
• computational economics.

A purely conventional market-share analysis may not adequately capture AI power.

22. Ex Ante and Ex Post Regulation

A major institutional question is whether competition authorities should intervene before monopolization becomes entrenched.

Ex post model

Intervene after:

• dominance;
• exclusion;
• foreclosure;
• consumer harm.

Ex ante model

Impose obligations upon strategically important AI platforms before harmful conduct occurs.

The latter approach may be particularly relevant where network effects and learning effects can rapidly produce irreversible market concentration.

23. International Competition Issues

AI markets are inherently cross-border.

A single AI ecosystem may involve:

U.S. model developer → Asian semiconductor supply → European users → global cloud infrastructure → Indian developers.

Consequently, competition authorities may need greater cooperation concerning:

• mergers;
• cross-border investigations;
• information exchange;
• remedies;
• multinational AI platforms.

Divergent national remedies could otherwise produce regulatory fragmentation.

24. Balancing Competition With Innovation

Competition law must distinguish between monopolization and successful innovation.

A firm should not be penalized merely because it develops a superior AI system.

The important distinction is between:

obtaining market power through innovation

and

using market power to prevent effective competition.

This distinction is especially important because aggressive intervention could potentially reduce incentives to invest in expensive AI research.

25. Future Intelligence Monopoly: A Conceptual Framework

The development can be represented as follows:

Compute Advantage
↓
Training Advantage
↓
Superior Model
↓
More Users
↓
More Data + Developer Adoption
↓
Better Ecosystem
↓
Higher Switching Costs
↓
Greater Distribution Power
↓
Entry Barriers
↓
Potential Intelligence Monopolization

Competition law must determine at which stage intervention is justified.

26. Key Legal Issues for Future Research

The most significant emerging questions include:

1. Can a foundation model constitute a relevant antitrust market?
2. When does control over AI compute constitute market power?
3. Can proprietary training data constitute an essential facility?
4. When should AI interoperability be mandatory?
5. Can autonomous AI agents participate in unlawful coordination?
6. Should algorithmic pricing systems create liability for their users?
7. How should AI acquisitions be assessed where the target has little revenue?
8. Can an AI platform lawfully privilege its own agents?
9. How should competition law address AI-cloud vertical integration?
10. Can competition authorities impose structural separation on AI ecosystems?
11. How should intellectual-property rights interact with access to critical AI technologies?
12. What constitutes exclusionary conduct in an autonomous economy?

Conclusion

Future intelligence monopolization represents a potential evolution from product-based monopoly to ecosystem-based control of economic intelligence.

The central competition-law challenge will not simply be determining which firm has the largest AI model. It will involve examining control over the entire chain of intelligence production and distribution:

Semiconductors → Compute → Data → Models → Agents → Applications → Distribution → Users.

The principles developed in Microsoft, Google Shopping, Google Android, Bronner, Magill, IMS Health, Intel and Qualcomm provide important foundations for addressing these problems. Their application to AI will require careful attention to network effects, data advantages, interoperability, vertical integration, innovation incentives and the possibility that AI ecosystems can become self-reinforcing.