Competition Law and Governance of Advanced Intelligence-Driven Market Systems

Introduction

Advanced intelligence-driven market systems are markets in which artificial intelligence (AI), machine learning, autonomous agents, predictive analytics, algorithmic pricing, recommendation engines, large-scale data systems, foundation models, automated decision-making and digital ecosystems substantially influence competitive conditions.

These systems can improve efficiency, reduce transaction costs and facilitate innovation. At the same time, they can create new forms of market power because the entity controlling the data, computational infrastructure, algorithms, interfaces, distribution channels or AI ecosystem may be able to influence how markets operate.

Traditional competition law therefore increasingly has to examine not merely whether a firm has a large market share, but also whether it controls an intelligence infrastructure that competitors cannot realistically replicate.

The principal competition-law concerns include:

1. algorithmic collusion;
2. AI-enabled abuse of dominance;
3. control over data and compute;
4. self-preferencing by intelligent platforms;
5. AI-driven exclusionary pricing;
6. interoperability restrictions;
7. tying and bundling of AI services;
8. discriminatory access to AI infrastructure;
9. acquisitions of emerging AI competitors;
10. concentration of innovation and knowledge;
11. autonomous commercial agents;
12. manipulation of recommendation and ranking systems.

I. Meaning of Intelligence-Driven Market Systems

An intelligence-driven market system is one in which commercial outcomes are significantly determined by computational intelligence rather than solely by conventional human decision-making.

Examples include:

• AI-powered search engines;
• algorithmic marketplaces;
• autonomous pricing systems;
• AI recommendation platforms;
• digital advertising exchanges;
• AI-powered financial markets;
• autonomous logistics systems;
• cloud-based AI infrastructure;
• foundation-model ecosystems;
• AI-enabled healthcare platforms;
• intelligent energy-management systems;
• autonomous vehicle platforms;
• AI procurement systems.

The competitive structure can be represented as:

Data → Compute → Model → Algorithm → Platform → Distribution → Consumers

Control over several stages simultaneously can generate substantial competitive advantages.

II. Competition-Law Framework

1. Market Definition

Traditional market definition remains relevant, but AI markets frequently involve:

• zero-price services;
• multi-sided platforms;
• rapidly changing technology;
• non-price competition;
• data-driven competition;
• innovation competition.

For example, a search engine may provide services to consumers at zero monetary price while simultaneously operating an advertising market.

Competition authorities therefore may need to consider:

• quality;
• privacy;
• innovation;
• data access;
• switching costs;
• interoperability;
• network effects;
• attention;
• algorithmic visibility.

III. Sources of Market Power in AI Markets

A. Data Advantages

Large datasets can produce:

More data → better model → more users → more data

This creates a feedback loop.

Where access to data is indispensable for competing effectively, competition authorities may examine whether dominant firms unlawfully restrict access.

B. Compute Power

Advanced AI requires substantial:

• GPUs;
• cloud computing;
• specialised chips;
• data centres;
• energy;
• technical infrastructure.

Control over scarce computational infrastructure can therefore become a source of market power.

A vertically integrated firm controlling:

Cloud → Chips → Foundation Model → Applications

may have incentives and opportunities to disadvantage rival AI developers.

C. Network Effects

AI platforms may become stronger as more users interact with them.

For example:

More users → more interactions → more training data → better AI → more users

Such feedback mechanisms can produce rapid concentration.

D. Switching Costs

Users may become dependent on:

• proprietary APIs;
• stored prompts;
• customised models;
• enterprise integrations;
• proprietary data formats;
• cloud infrastructure;
• workflow automation.

High switching costs may reinforce incumbent power.

IV. Algorithmic Collusion

One of the most important future competition concerns is algorithmic coordination.

Two competing firms may independently deploy pricing algorithms that continuously observe market prices and adjust their own prices.

Even without an explicit human agreement, algorithms may facilitate:

• parallel pricing;
• rapid price matching;
• market allocation;
• output restriction;
• retaliation against discounting.

Competition law must distinguish between:

legitimate intelligent adaptation and algorithmic coordination that facilitates prohibited concerted conduct.

V. Case Law

1. Eturas UAB v Lietuvos Respublikos konkurencijos taryba — C-74/14

The Court of Justice of the European Union considered an electronic travel-booking platform where a system message effectively restricted discounts that travel agencies could provide.

The case is important because competition law can apply where an electronic platform facilitates coordination among otherwise independent businesses.

Principle

A digital system cannot be used as a mechanism for coordinating competitors merely because the coordination is technologically implemented.

Relevance to AI

An AI platform capable of communicating pricing recommendations among competing businesses could create similar concerns.

The key question becomes whether participants knowingly participate in or facilitate coordinated conduct.

2. United States v. Google LLC — Search and Search Advertising

The U.S. Google search litigation examined Google's conduct concerning distribution arrangements and default-search positions.

The competition concerns included Google's ability to reinforce its position through agreements affecting access to important distribution channels.

Principle

Dominance can be strengthened through control over distribution and default access, not merely through ownership of the underlying technology.

Relevance to AI

An AI assistant integrated as the default service in:

• operating systems;
• browsers;
• smartphones;
• cloud environments;

could potentially reinforce the provider's position in adjacent AI markets.

3. Google Shopping — Case C-48/22 P

The European Court of Justice upheld the central finding concerning Google's preferential positioning of its own comparison-shopping service in search results.

The case is especially relevant to self-preferencing.

Principle

A dominant digital platform may face competition-law scrutiny where it uses control over an important platform or infrastructure to favour its own downstream service.

Relevance to AI

An AI platform could potentially:

• favour its own AI applications;
• prioritise its own agents;
• promote its own shopping services;
• privilege its own advertising products;
• manipulate recommendation results.

The central competition question is whether control over the platform is being used to distort competition in adjacent markets.

4. Google Android — Case T-604/18

The EU General Court examined Google's contractual arrangements concerning Android devices, including restrictions associated with search and browser distribution.

The case illustrates how dominance in one technological layer can be leveraged into neighbouring markets.

Principle

Competition law may address contractual arrangements that reinforce dominance across interconnected technological ecosystems.

AI relevance

Similar issues may arise where a company controlling an AI operating layer conditions access to:

• APIs;
• cloud services;
• application stores;
• AI models;
• operating systems.

An AI ecosystem can therefore create cross-market leveraging.

5. Intel v European Commission — Case C-413/14 P

The Intel litigation concerned rebates offered by a dominant undertaking and the assessment of exclusionary effects.

The Court of Justice emphasised the importance of examining the actual or potential ability of conduct to foreclose equally efficient competitors.

Principle

The analysis of exclusionary conduct cannot always be reduced to the formal classification of a practice; its competitive effects may require careful examination.

AI relevance

AI companies may offer:

• cloud discounts;
• bundled AI credits;
• preferential API pricing;
• volume rebates;
• exclusivity incentives.

Competition authorities may therefore need to examine whether such arrangements exclude rival AI providers.

6. Slovak Telekom and Deutsche Telekom — Joined Cases C-152/19 P and C-165/19 P

The litigation concerned exclusionary conduct involving access to telecommunications infrastructure.

It demonstrates the importance of analysing access to infrastructure controlled by a dominant undertaking.

Principle

Where a dominant undertaking controls infrastructure essential to downstream competition, restrictions on access can raise abuse-of-dominance concerns.

AI relevance

The same conceptual issue may arise with:

• AI compute;
• cloud infrastructure;
• proprietary AI interfaces;
• critical datasets;
• AI model access.

The legal analysis would still require satisfaction of the applicable legal tests; technological importance alone does not automatically make an infrastructure facility legally indispensable.

7. Google AdSense — Case T-334/19

The EU litigation concerning Google's AdSense practices addressed contractual restrictions affecting competition in online advertising.

Principle

Restrictions imposed through commercial contracts can become competition concerns where they restrict rivals' ability to compete effectively.

AI relevance

AI advertising ecosystems may involve contracts concerning:

• data access;
• model access;
• advertising inventory;
• interoperability;
• API integration;
• exclusive use of AI advertising tools.

Thus, intelligent advertising systems should not be assessed only through their technical architecture; their contractual governance is equally important.

8. Microsoft / Commission Competition Cases

The Microsoft litigation concerning tying and interoperability remains particularly significant for technology markets.

The disputes demonstrated that control over a dominant software platform can be used to influence competition in adjacent technological markets.

Principle

A dominant technological ecosystem can create competitive problems where interoperability or access restrictions prevent rivals from competing effectively.

AI relevance

The same concern can arise when a dominant AI ecosystem controls:

model + operating system + cloud + applications + data + APIs.

AI competition may therefore increasingly resemble earlier platform and operating-system competition disputes.

VI. AI Self-Preferencing

AI systems introduce a new form of self-preferencing.

A traditional search engine might place its own product first.

An AI assistant can go further by determining:

• which products are recommended;
• which suppliers are visible;
• which sources are summarised;
• which services are called through an agent;
• which merchants receive transactions.

The AI system therefore becomes not merely an information intermediary but potentially a market-making intermediary.

VII. Autonomous AI Agents and Competition Law

A particularly difficult future problem is the emergence of autonomous commercial agents.

Suppose:

• Company A's AI agent negotiates prices;
• Company B's AI agent does the same;
• both agents continuously observe one another;
• both learn that aggressive price reductions reduce profitability;
• both independently converge on higher prices.

There may be no traditional human meeting or communication.

Competition law must therefore determine:

1. whether there was an agreement;
2. whether there was conscious coordination;
3. whether the firms designed the systems in a manner facilitating coordination;
4. whether the algorithms merely responded independently to market conditions;
5. who bears legal responsibility for autonomous conduct.

VIII. AI and Abuse of Dominance

Potential exclusionary practices include:

1. Refusal to provide API access

A dominant platform may deny competitors access to an interface necessary to interoperate with its ecosystem.

2. Discriminatory access

The platform may provide superior AI functionality to its own products while offering inferior access to competitors.

3. Tying

A dominant cloud provider might require customers purchasing one AI service to purchase another service.

4. Bundling

AI functionality may be bundled with:

• cloud hosting;
• operating systems;
• productivity software;
• cybersecurity;
• advertising;
• enterprise applications.

5. Predatory pricing

A dominant firm might temporarily price AI services below sustainable levels to eliminate smaller rivals.

6. Margin squeeze

A vertically integrated AI company could charge rivals high prices for infrastructure while competing against them downstream at low prices.

IX. Data as a Competition Resource

Data can generate competitive advantages through:

• scale;
• variety;
• quality;
• real-time access;
• feedback loops.

However, possession of large quantities of data does not automatically establish dominance.

Authorities must examine:

• whether the data are unique;
• whether substitutes exist;
• whether competitors can obtain equivalent data;
• whether the data materially improve the relevant product;
• whether access restrictions exclude competitors.

X. AI Mergers and Acquisitions

AI markets create new merger-control questions.

Traditional merger analysis focuses on:

• market shares;
• concentration;
• competitive overlaps;
• barriers to entry.

AI requires additional attention to:

• datasets;
• computing capacity;
• engineering talent;
• foundation models;
• patents;
• APIs;
• developer ecosystems;
• distribution;
• strategic partnerships.

A transaction involving a small AI company may still raise competition concerns if the target is an important potential innovator.

XI. Killer Acquisitions in AI

A large incumbent may acquire a young AI company before it becomes a significant competitor.

Competition authorities may therefore examine:

Current competition + potential competition + innovation competition

rather than simply asking whether the target currently has substantial revenue.

Relevant factors may include:

• research pipeline;
• technological capabilities;
• customer adoption;
• venture funding;
• likelihood of independent expansion;
• access to critical data;
• ability to challenge incumbents.

XII. Competition in AI Foundation Models

Foundation models can function as upstream infrastructure.

Potential market structure:

Compute Provider

↓

Foundation Model

↓

API

↓

AI Applications

↓

Consumers / Businesses

If one undertaking controls multiple layers, vertical foreclosure becomes a major competition concern.

For example, a foundation-model provider might theoretically favour its own downstream application by:

• giving it preferential API access;
• providing better model performance;
• imposing discriminatory terms on competitors;
• restricting interoperability.

XIII. Interoperability

Interoperability can become a major competition remedy.

Possible measures include:

• API access;
• data portability;
• model interoperability;
• technical standards;
• switching tools;
• open interfaces;
• portability of enterprise AI configurations.

Interoperability may reduce switching costs and prevent ecosystem lock-in.

However, mandatory access can also affect:

• cybersecurity;
• intellectual property;
• privacy;
• model safety;
• trade secrets.

Competition remedies must therefore be carefully designed.

XIV. Algorithmic Discrimination

AI systems may rank businesses differently based on:

• historical data;
• conversion rates;
• consumer behaviour;
• advertising expenditure;
• platform relationships.

This raises competition concerns where a dominant platform's algorithm systematically disadvantages rival businesses.

The distinction between legitimate algorithmic optimisation and unlawful discriminatory conduct requires analysis of:

• objective;
• effects;
• transparency;
• economic rationale;
• alternatives;
• competitive impact.

XV. Competition and Innovation

Traditional competition law often focuses on price and output.

AI markets require greater emphasis on:

Innovation competition

A firm may harm competition even without immediately raising prices if it:

• suppresses rival innovation;
• acquires emerging technologies;
• prevents interoperability;
• restricts access to critical datasets;
• forecloses alternative AI architectures.

The relevant competitive harm may therefore occur before conventional price effects become visible.

XVI. Governance of Intelligent Market Systems

Competition law should increasingly be viewed as part of a broader governance architecture.

A useful framework is:

Layer 1 — Infrastructure Governance

• cloud;
• chips;
• data centres;
• networks.

Layer 2 — Data Governance

• data access;
• portability;
• interoperability;
• privacy.

Layer 3 — Model Governance

• foundation models;
• APIs;
• model access.

Layer 4 — Platform Governance

• ranking;
• recommendation;
• app stores;
• marketplaces.

Layer 5 — Agent Governance

• autonomous purchasing;
• autonomous pricing;
• autonomous negotiation.

Layer 6 — Consumer Governance

• transparency;
• choice;
• switching;
• protection against manipulation.

Competition law intersects with each layer.

XVII. Future Competition-Law Tests

Advanced intelligence-driven markets may require competition authorities to examine five dimensions simultaneously:

This represents a movement from market-share analysis toward infrastructure-and-ecosystem analysis, while retaining conventional economic and legal tests.

XVIII. Remedies

Potential competition remedies include:

Structural remedies

• divestiture;
• separation of business units;
• limits on acquisitions.

Behavioural remedies

• non-discrimination;
• fair-access obligations;
• restrictions on exclusivity;
• transparency requirements.

Interoperability remedies

• API access;
• data portability;
• technical interoperability.

Algorithmic remedies

• independent auditing;
• monitoring;
• restrictions on discriminatory ranking;
• preservation of decision records.

Merger remedies

• licensing;
• access commitments;
• divestiture of overlapping assets;
• restrictions on exclusive arrangements.

XIX. Key Legal Challenges

1. Attribution

Who is responsible when an autonomous AI system makes an anticompetitive decision?

2. Intent

Traditional competition law sometimes relies on evidence of knowledge or intention. AI systems complicate this because outcomes may emerge from machine learning.

3. Causation

It can be difficult to demonstrate that an AI system caused competitive harm rather than merely responding to market conditions.

4. Explainability

Authorities may need to understand why an algorithm:

• increased prices;
• demoted a rival;
• recommended a particular product;
• denied access.

5. Dynamic markets

AI markets can change rapidly, making conventional market definition potentially temporary.

6. Innovation uncertainty

Authorities must avoid protecting inefficient competitors merely because they are smaller while also preventing dominant firms from eliminating genuine future competitors.

XX. A Proposed Intelligence-Competition Governance Model

A future framework can be represented as:

Identify Critical AI Infrastructure

↓

Define Relevant Markets and Ecosystems

↓

Measure Data + Compute + Network Effects

↓

Identify Dominant/Strategically Important Actors

↓

Examine Conduct

↓

Assess Exclusionary or Coordinating Effects

↓

Assess Innovation Effects

↓

Examine Interoperability and Access

↓

Consider Merger/Acquisition Risks

↓

Apply Proportionate Remedies

↓

Continuous Algorithmic Monitoring

This final stage is particularly important because AI systems can change their behaviour after deployment.

XXI. Overall Legal Significance

The cases discussed above demonstrate that competition law already contains several principles capable of addressing AI markets:

• Eturas illustrates electronic facilitation of coordination.
• Google Shopping addresses self-preferencing concerns.
• Google Android demonstrates ecosystem leveraging.
• Intel illustrates effects-based analysis of exclusionary conduct.
• Slovak Telekom demonstrates the significance of access to infrastructure.
• Google AdSense illustrates contractual restrictions in digital markets.
• Microsoft-related competition jurisprudence demonstrates the importance of interoperability and platform leverage.
• Google search litigation in the United States illustrates the significance of distribution and default arrangements.

The challenge is therefore not necessarily to replace competition law with an entirely new legal system. Rather, competition authorities may need to adapt established principles to markets in which intelligence, data, compute and autonomous decision-making themselves become sources of market power.

Conclusion

Competition law in advanced intelligence-driven market systems must move beyond the simple question of who sells the most products.

The critical questions increasingly become:

Who controls the data?

Who controls the compute?

Who controls the models?

Who controls access to users?

Who controls the algorithms that determine commercial visibility?

Who controls interoperability?

Who can prevent competitors from reaching the next stage of technological development?

The central competition-law challenge is to preserve contestable, innovative and interoperable markets while allowing firms to obtain legitimate returns from investment in AI, data and infrastructure.

Accordingly, future competition governance is likely to combine traditional doctrines of dominance, exclusion, tying, refusal of access, vertical restraints, merger control and cartel prohibition with closer scrutiny of data concentration, algorithmic coordination, AI ecosystems, autonomous agents, compute dependence and innovation foreclosure.