Competition Law and Future-Oriented Competition Governance for Advanced Intelligent Systems

1. Introduction

Future-oriented competition governance for advanced intelligent systems concerns the application and development of competition law where markets are increasingly shaped by artificial intelligence (AI), autonomous decision-making, foundation models, algorithmic pricing, data-intensive ecosystems, machine-learning infrastructure, and intelligent platforms.

Traditional competition law generally asks whether a firm has market power, whether competitors have entered into an anticompetitive agreement, whether a dominant undertaking has abused its position, or whether a merger substantially reduces competition. Advanced intelligent systems complicate these questions because competitive advantages may arise from data, computing capacity, algorithms, model quality, ecosystems, interoperability, access to interfaces, and control over distribution channels rather than conventional physical assets.

A future-oriented governance framework therefore examines not only existing market shares but also:

• control over strategically important data;
• access to computing infrastructure;
• AI-model interoperability;
• algorithmic coordination;
• self-preferencing by intelligent platforms;
• exclusionary model licensing;
• acquisition of emerging AI competitors;
• access to application programming interfaces (APIs);
• switching and migration barriers;
• network and learning effects;
• vertical integration between AI infrastructure and applications;
• transparency and auditability of algorithmic decisions; and
• the possibility that today's apparently small AI firm could become an important future competitive constraint.

2. Meaning of Advanced Intelligent Systems

The expression advanced intelligent systems can encompass:

1. Foundation models – large general-purpose AI models capable of supporting multiple applications.
2. Generative AI systems – systems producing text, images, audio, video or code.
3. Autonomous agents – systems capable of planning and executing tasks with limited human intervention.
4. Intelligent recommendation systems – algorithms determining what consumers see, purchase or access.
5. Algorithmic pricing systems – automated systems capable of setting or adjusting prices.
6. AI-powered marketplaces – platforms matching consumers and suppliers through machine learning.
7. Autonomous vehicles and robotics.
8. AI-enabled cloud infrastructure.
9. Decision-support systems in finance, healthcare, logistics and other sectors.
10. Multi-layer AI ecosystems involving chips, cloud computing, models, applications and distribution.

Competition law therefore increasingly operates across an AI value chain, rather than merely within an individual product market.

3. Why Future-Oriented Competition Governance Is Necessary

A. Rapid technological development

AI markets can develop considerably faster than traditional regulatory processes.

A company may move from:

research → model development → platform integration → ecosystem dominance

within a comparatively short period.

Consequently, competition authorities may need to identify competitive threats before exclusion becomes irreversible.

B. Data-driven market power

Data can create competitive advantages through:

• training data;
• user interaction data;
• behavioural information;
• transaction data;
• search histories;
• feedback data;
• model-performance data.

The important competition question is not simply whether a firm possesses data, but whether exclusive control over strategically important data materially impedes competitors.

C. Computing infrastructure

Advanced AI frequently requires substantial computational resources.

Relevant inputs may include:

• GPUs;
• AI accelerators;
• cloud computing;
• specialised data centres;
• inference infrastructure;
• model-hosting services.

Competition problems may therefore arise when an undertaking controls several vertically related levels of the AI stack.

4. Competition Layers in Intelligent Markets

A useful analytical model is:

Semiconductors → Computing → Cloud → Foundation Models → AI Applications → Distribution → Consumers

Competition problems can arise at every level.

For example:

Infrastructure layer

Possible concerns:

• exclusive supply;
• capacity reservation;
• discriminatory access;
• tying;
• refusal to supply.

Model layer

Potential concerns include:

• exclusive model licensing;
• interoperability restrictions;
• discriminatory API access;
• restrictions on model portability.

Application layer

Potential issues include:

• self-preferencing;
• tying;
• bundling;
• exclusionary defaults;
• discriminatory ranking.

Distribution layer

Possible concerns include:

• app-store restrictions;
• search ranking;
• default placement;
• access restrictions;
• platform fees.

5. Traditional Competition Law Doctrines Applied to AI

A. Article 101 TFEU / Section 1 Sherman Act / Competition Act equivalents

Agreements between AI firms can raise concerns where they involve:

• price coordination;
• output restrictions;
• market allocation;
• information exchange;
• customer allocation;
• exclusionary licensing.

The fact that coordination is achieved through algorithms does not automatically remove it from competition law.

B. Abuse of dominance

A dominant AI platform may potentially engage in:

• discriminatory access;
• tying;
• bundling;
• self-preferencing;
• refusal to interoperate;
• exclusionary rebates;
• predatory pricing;
• discriminatory algorithmic ranking.

The central issue remains whether the conduct harms competition rather than merely whether an algorithm is sophisticated.

C. Merger control

AI creates particularly important merger questions.

Authorities may need to examine acquisitions involving:

• startups with low current revenues;
• valuable AI researchers;
• critical datasets;
• proprietary models;
• AI infrastructure;
• emerging competitors.

Traditional turnover thresholds may fail to capture transactions involving firms whose current revenue is low but competitive significance is potentially high.

6. Six Important Case Laws

1. Google Search (Shopping) – Google LLC v European Commission

European Commission decision, General Court, Case T-612/17; CJEU appeal, Case C-48/22 P

Principle

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

The competition concern involved Google's use of its dominant general-search position to give favourable treatment to its own comparison-shopping service.

Relevance to intelligent systems

The case provides an important conceptual foundation for future AI markets because intelligent assistants and AI search systems may control the ranking and visibility of competing services.

An AI assistant could potentially:

• recommend its own service;
• rank affiliated products preferentially;
• suppress competing results;
• control consumer discovery.

The key future-oriented question becomes whether an AI intermediary is merely providing an improved service or using an entrenched position to foreclose competing providers.

7. Google Android – Google and Alphabet v European Commission

General Court, Case T-604/18; Commission Android decision

Principle

The Android case concerned Google's practices involving licensing and distribution arrangements for its mobile operating system and related services.

The case examined practices involving:

• tying;
• pre-installation;
• distribution incentives;
• restrictions affecting competing search services.

AI relevance

The case is particularly relevant to future intelligent ecosystems because AI assistants may become the principal interface through which consumers interact with digital services.

A dominant ecosystem could potentially require:

AI assistant + search + browser + payments + cloud + applications

as a bundled environment.

Competition governance may therefore need to assess whether intelligent interfaces create new forms of ecosystem foreclosure.

8. Microsoft – Interoperability and Tying

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

Principle

The Microsoft litigation addressed Microsoft's dominant position in operating systems and conduct involving interoperability information and the tying of products.

The case is important for the proposition that competition law may intervene where a dominant undertaking uses control over one technological layer to disadvantage competition in another.

AI relevance

Advanced intelligent ecosystems can exhibit similar vertical relationships:

operating system → cloud → AI model → AI assistant → applications

If a dominant infrastructure provider restricts interoperability with competing AI systems, questions may arise concerning:

• access to technical interfaces;
• model portability;
• interoperability;
• compatibility;
• data access.

The Microsoft reasoning therefore offers a useful framework for examining future AI ecosystems.

9. Intel v Commission

Case C-413/14 P

Principle

The Intel litigation concerned rebates provided by a dominant undertaking and the assessment of whether such conduct could produce exclusionary effects.

The later judgment emphasised the importance of examining economic effects where the undertaking concerned provides evidence relevant to the competitive assessment.

AI relevance

AI infrastructure markets can involve powerful suppliers of:

• processors;
• GPUs;
• accelerators;
• cloud infrastructure.

Long-term contracts, capacity commitments or loyalty arrangements could potentially affect competitors' ability to obtain critical computing resources.

The Intel jurisprudence therefore helps inform analysis of exclusionary contractual practices in AI infrastructure markets.

10. Qualcomm v Commission

Case T-235/18

Principle

The Qualcomm litigation concerned payments and exclusivity arrangements involving baseband chipsets.

The case demonstrates the importance of examining contractual arrangements that may affect access to important technology markets.

AI relevance

Comparable issues could emerge in advanced AI infrastructure where an undertaking provides incentives for customers or downstream firms to use its AI technology exclusively.

Potential competition concerns could involve:

• exclusive AI-chip arrangements;
• exclusive cloud commitments;
• model exclusivity;
• bundled infrastructure;
• preferential access agreements.

The case therefore provides an analytical bridge between traditional technology competition law and future AI infrastructure markets.

11. United States v. Microsoft Corp.

253 F.3d 34 (D.C. Cir. 2001)

Principle

The Microsoft case examined Microsoft's conduct concerning the Windows operating-system monopoly and competing browsers.

The court considered practices that could preserve or strengthen market power by limiting competing distribution opportunities.

AI relevance

The case has substantial conceptual relevance to AI because distribution can be as important as technological superiority.

An AI developer might have a technically competitive model but face exclusion if a dominant platform controls:

• default AI assistants;
• operating-system access;
• app distribution;
• search interfaces;
• cloud deployment;
• consumer devices.

Future competition governance must therefore consider whether control over an interface becomes a bottleneck for AI competition.

12. United States v. Google – Search and Search Advertising

United States v Google LLC, U.S. District Court for the District of Columbia

Principle

The Google search litigation examined agreements and practices alleged to preserve Google's position in general search and search advertising.

AI relevance

The case is highly significant for future intelligent systems because generative AI increasingly changes how users discover information.

The competitive question may move from:

"Which search engine receives the query?"

to:

"Which AI system controls the answer, recommendation or transaction?"

This creates potential competition concerns surrounding:

• default AI placement;
• exclusive distribution;
• access to search data;
• advertising integration;
• AI-generated recommendations;
• downstream commercial transactions.

13. AI-Specific Future Competition Problems

A. Algorithmic collusion

AI systems can independently observe competitors' prices and adjust their own prices.

The difficult legal question is whether autonomous machine behaviour amounts to:

• conscious coordination;
• facilitating conduct;
• unlawful information exchange;
• independent parallel conduct.

Future competition law may therefore require methods for distinguishing independent algorithmic adaptation from coordinated conduct.

B. Algorithmic discrimination

Intelligent platforms may rank:

• sellers;
• products;
• advertisements;
• search results;
• applications

using complex models.

If the platform also participates in the market being ranked, self-preferencing concerns may arise.

14. AI Self-Preferencing

Consider an AI marketplace where the platform sells its own products.

The AI could systematically recommend:

Platform-owned product → affiliated seller → independent competitor.

Competition analysis should examine:

1. how the ranking algorithm works;
2. whether the platform has market power;
3. whether affiliated products receive systematic advantages;
4. whether competitors can obtain equivalent visibility;
5. whether consumers are misled;
6. whether exclusionary effects are likely.

15. Data as a Competition Asset

Data can generate:

Network effects

More users → more data → better AI → more users.

This can create a reinforcing feedback loop:

Users → Data → Model improvement → Better service → More users

If competitors cannot reproduce this loop, entry barriers may increase.

However, possession of large datasets should not automatically be equated with unlawful market power.

The relevant question is whether the data is:

• commercially significant;
• difficult to replicate;
• indispensable or strategically important;
• capable of producing competitive advantages.

16. AI and Essential Facilities

Traditional essential-facility reasoning may become relevant to:

• specialised computing;
• critical datasets;
• AI interfaces;
• model APIs;
• cloud infrastructure.

A refusal to provide access does not automatically violate competition law.

The legal assessment generally depends upon factors such as:

• indispensability;
• dominance;
• feasibility of duplication;
• competitive foreclosure;
• objective justification.

17. Interoperability as a Competition Remedy

Future competition governance may increasingly use interoperability requirements.

Possible remedies include:

API access

Competitors receive access under transparent and non-discriminatory conditions.

Data portability

Users can transfer relevant information between AI systems.

Model portability

Applications can switch between compatible AI models.

Interface interoperability

Different intelligent systems can communicate with one another.

These measures can reduce switching costs and prevent ecosystem lock-in.

18. AI Mergers and Killer Acquisitions

Traditional merger control may focus heavily on:

current turnover + existing market share.

AI requires consideration of additional factors:

• technological capabilities;
• patents;
• specialised researchers;
• training data;
• computing resources;
• developer ecosystems;
• future competitive potential;
• innovation pipelines.

A startup with modest revenue could nevertheless be strategically important because it represents a potential future competitive constraint.

19. Acqui-Hiring and AI Talent

AI firms may acquire companies primarily to obtain:

• researchers;
• engineers;
• model-development teams;
• specialised knowledge.

Competition authorities may therefore need to examine whether a transaction eliminates an emerging source of innovation even where the acquired firm's conventional market share is small.

20. Cloud–AI Vertical Integration

One of the most important future governance problems is vertical integration:

Cloud provider → computing infrastructure → foundation model → AI application → distribution

Vertical integration can create efficiencies because infrastructure and models may work better together.

But it can also create foreclosure risks if a vertically integrated undertaking:

• denies rivals computing capacity;
• offers preferential pricing to its own AI applications;
• restricts interoperability;
• bundles cloud and AI services;
• imposes exclusivity.

Competition law therefore needs to balance innovation efficiencies against foreclosure risks.

21. Intelligent Agents and Competition

Autonomous AI agents could eventually:

• negotiate prices;
• purchase goods;
• select suppliers;
• change contracts;
• switch platforms;
• optimise logistics.

This could generate both pro-competitive and anticompetitive consequences.

Pro-competitive effect

Agents could reduce search costs and increase consumer bargaining power.

Possible anticompetitive effect

If many agents use similar optimisation objectives or are controlled by a small number of providers, markets could become highly concentrated around a few algorithmic decision systems.

22. Competition Between AI Ecosystems

Future competition may not be between individual products but between ecosystems.

For example:

Ecosystem A

• chips
• cloud
• foundation model
• assistant
• app store

versus

Ecosystem B

• alternative infrastructure
• alternative model
• independent applications
• independent distribution.

Competition authorities may therefore need to analyse ecosystem-level market power.

23. Future-Oriented Governance Model

A comprehensive governance framework can be divided into six stages.

Stage 1 – Market Mapping

Identify:

• AI infrastructure;
• models;
• applications;
• distribution channels;
• data;
• complementary services.

Stage 2 – Market-Power Assessment

Examine:

• market shares;
• switching costs;
• network effects;
• data advantages;
• computational advantages;
• entry barriers.

Stage 3 – Conduct Assessment

Investigate:

• tying;
• bundling;
• exclusivity;
• self-preferencing;
• discriminatory access;
• predatory strategies;
• algorithmic coordination.

Stage 4 – Innovation Assessment

Examine:

• R&D competition;
• potential entrants;
• innovation pipelines;
• access to talent;
• technological alternatives.

Stage 5 – Ecosystem Assessment

Analyse:

• interoperability;
• portability;
• APIs;
• vertical integration;
• platform dependence.

Stage 6 – Remedies

Possible remedies include:

• behavioural commitments;
• interoperability;
• data portability;
• non-discrimination;
• divestiture;
• licensing;
• access obligations;
• structural separation in exceptional cases.

24. Role of Ex Ante Regulation

Traditional competition law is primarily ex post.

Future intelligent markets may also require ex ante safeguards, particularly where:

• network effects are extremely strong;
• switching costs are substantial;
• markets tip rapidly;
• dominant platforms control essential interfaces.

The European Union's Digital Markets Act illustrates this movement toward predefined obligations for designated gatekeepers.

The objective is not to replace competition law but to supplement it where conventional enforcement may occur too late.

25. Relationship Between Competition Law and AI Regulation

Competition law should operate alongside:

• data protection;
• AI regulation;
• consumer protection;
• cybersecurity;
• intellectual property law;
• sector-specific regulation.

These regimes have different objectives.

Future governance requires coordination rather than treating these regimes as isolated systems.

26. Regulatory Challenges

A. Defining the relevant market

Traditional product boundaries may become unstable when one AI system performs many functions.

A single assistant might simultaneously operate as:

• search engine;
• translator;
• coding tool;
• shopping intermediary;
• productivity application;
• recommendation engine.

B. Measuring market power

Market share may not adequately capture:

• data advantages;
• computational capacity;
• model quality;
• ecosystem effects;
• developer dependence.

C. Proving algorithmic causation

Complex machine-learning systems can make it difficult to determine precisely why an exclusionary outcome occurred.

Competition authorities may therefore need:

• technical audits;
• algorithmic evidence;
• model documentation;
• data-access records;
• API logs;
• experimentation evidence.

27. Compliance Framework for AI Companies

An AI company should establish:

1. Competition-law risk assessment

Regularly review:

• pricing algorithms;
• distribution agreements;
• API policies;
• exclusivity;
• bundling.

2. Algorithm governance

Maintain records concerning:

• objectives;
• constraints;
• training;
• deployment;
• changes to ranking systems.

3. Merger compliance

Assess acquisitions for:

• horizontal overlap;
• vertical foreclosure;
• innovation competition;
• emerging competitors.

4. Interoperability policy

Document objective reasons for:

• API restrictions;
• access limitations;
• compatibility requirements.

5. Data governance

Identify whether data practices:

• foreclose competitors;
• discriminate between users;
• create artificial switching costs.

28. Future Competition Governance Principles

A coherent future-oriented system should incorporate the following principles:

Principle 1 – Competitive neutrality

AI systems should not use control over one market to unfairly foreclose competition in another.

Principle 2 – Interoperability

Where technologically and legally appropriate, interoperability can reduce ecosystem lock-in.

Principle 3 – Innovation preservation

Competition law should protect the possibility of future technological competition, not merely current competitors.

Principle 4 – Algorithmic accountability

Competition authorities must be able to understand how commercially significant algorithms affect competitive conditions.

Principle 5 – Data accessibility

Data advantages should be examined where exclusive control creates significant competitive barriers.

Principle 6 – Dynamic market assessment

Authorities should consider how markets may evolve rather than relying exclusively on historical market shares.

Principle 7 – Proportionality

Intervention should distinguish legitimate technological integration from exclusionary conduct.

29. Overall Legal Framework

The future-oriented competition framework can therefore be represented as:

AI Infrastructure
↓
Computing & Cloud
↓
Foundation Models
↓
AI Applications
↓
Intelligent Distribution
↓
Consumers and Businesses

At every stage, competition law asks:

Market Power → Conduct → Effects → Justification → Remedy

This should be supplemented by:

Data + Algorithms + Interoperability + Innovation + Ecosystem Effects

30. Conclusion

Future-oriented competition governance for advanced intelligent systems represents an evolution from conventional market regulation toward dynamic, technology-sensitive competition governance.

The central challenge is not simply determining whether an AI company is large. It is determining whether control over data, computing, models, interfaces, algorithms, distribution and ecosystems can be used to reduce actual or potential competition.

The jurisprudence of Microsoft, Google Shopping, Google Android, Intel and Qualcomm, together with the U.S. Microsoft and Google litigation, provides important foundations for analysing these problems. They demonstrate that competition law can address technological bottlenecks, tying, exclusivity, interoperability, distribution advantages and the leveraging of market power across connected markets.

For advanced intelligent systems, future competition governance is therefore likely to depend upon a combination of:

traditional antitrust + merger control + interoperability + data governance + algorithmic oversight + innovation protection + ex ante digital regulation.

The central legal objective remains preservation of a competitive process while allowing technological integration and innovation to develop where those efficiencies do not produce unjustified exclusion of competing technologies.