Competition Law and Competition Governance in Hyper-Intelligent Economies

1. Introduction

A hyper-intelligent economy is an economic environment in which artificial intelligence, autonomous agents, foundation models, predictive analytics, machine-learning systems, synthetic data, robotics, automated decision-making and algorithmic infrastructure play a central role in production, distribution, pricing, investment and consumer interaction.

In such an economy, competition is no longer determined solely by conventional factors such as price, output and market share. Competitive advantage may instead depend upon:

• access to high-quality data;
• computing capacity and AI chips;
• foundation models and model weights;
• cloud infrastructure;
• proprietary algorithms;
• training datasets;
• AI talent;
• interoperability;
• API access;
• autonomous purchasing agents;
• recommendation and ranking systems;
• intellectual-property rights;
• ecosystems and digital platforms; and
• control over technological standards.

Competition law therefore has to evolve from merely regulating human commercial conduct to regulating markets in which important competitive decisions may be made or implemented by machines.

The central governance question becomes:

How should competition law preserve rivalry when intelligence itself becomes an important economic infrastructure?

2. Meaning of a Hyper-Intelligent Economy

A hyper-intelligent economy goes beyond an ordinary digital economy.

Digital economy

Businesses use:

• websites;
• platforms;
• cloud computing;
• databases;
• online marketplaces; and
• digital payments.

AI economy

Businesses additionally use:

• machine learning;
• predictive analytics;
• generative AI;
• automated pricing;
• recommendation engines; and
• autonomous decision systems.

Hyper-intelligent economy

The system increasingly allows AI to:

1. analyse markets;
2. predict consumer behaviour;
3. determine prices;
4. negotiate transactions;
5. allocate resources;
6. select suppliers;
7. recommend products;
8. optimise logistics;
9. create new products;
10. interact with other AI systems; and
11. autonomously execute commercial decisions.

Consequently, competition governance must address machine-mediated competition.

3. Competition-Law Framework

Traditional competition law generally addresses four principal areas:

A. Anti-competitive agreements

Conduct involving:

• price fixing;
• market allocation;
• output restrictions;
• bid rigging;
• information exchange; and
• coordinated conduct.

B. Abuse of dominance

A powerful undertaking may engage in:

• exclusionary conduct;
• discriminatory access;
• tying;
• bundling;
• self-preferencing;
• refusal to deal;
• predatory pricing;
• margin squeezing; or
• discriminatory ranking.

C. Merger control

Competition authorities must assess:

• acquisitions of AI companies;
• acquisition of AI talent;
• acquisition of datasets;
• cloud/AI integration;
• vertical AI mergers;
• killer acquisitions; and
• ecosystem consolidation.

D. Competition governance

Modern competition governance increasingly includes:

• interoperability requirements;
• data portability;
• transparency;
• access obligations;
• algorithmic auditing;
• structural separation;
• non-discrimination;
• technical standards; and
• continuous regulatory monitoring.

The European Union's Digital Markets Act illustrates this transition from purely ex-post antitrust enforcement toward ex-ante governance of powerful digital gatekeepers. In July 2026, the European Commission imposed two DMA fines on Google totalling €890 million concerning self-preferencing and anti-steering.

4. Why Hyper-Intelligent Markets Create New Competition Problems

A. Data concentration

AI systems require enormous quantities of data.

A dominant undertaking may simultaneously control:

users → data → algorithms → AI model → better service → more users → more data.

This creates a data-feedback loop.

The competitive concern is not simply ownership of data. It is whether control over data creates an enduring competitive advantage that rivals cannot reasonably reproduce.

B. Computational concentration

Modern AI requires substantial:

• GPUs;
• AI accelerators;
• cloud infrastructure;
• electricity;
• specialised data centres; and
• networking capacity.

If a small number of undertakings control these inputs, downstream AI competitors may become dependent upon them.

This produces a possible compute bottleneck.

C. Foundation-model concentration

A foundation model can become an essential technological layer for:

• search;
• productivity software;
• financial services;
• healthcare;
• education;
• autonomous vehicles;
• robotics;
• advertising; and
• commerce.

A dominant foundation-model provider could potentially extend market power into downstream markets.

5. Algorithmic Self-Preferencing

Hyper-intelligent platforms may use algorithms to determine:

• ranking;
• recommendations;
• search results;
• advertising;
• product visibility;
• access to consumers.

If the platform owns competing products, it can potentially design its algorithm to favour its own services.

This issue is already illustrated by Google Shopping. The European Commission found that Google favoured its own comparison-shopping service in general search results. The General Court upheld the central finding of an abuse of dominance while modifying part of the Commission's reasoning concerning certain markets.

In a hyper-intelligent economy, self-preferencing could become much more sophisticated because an AI system could dynamically change ranking according to millions of variables.

6. Algorithmic Collusion

AI systems can independently observe:

• competitors' prices;
• demand;
• inventory;
• market conditions; and
• consumer responses.

They can then adjust prices automatically.

This creates a difficult legal distinction between:

Independent parallel behaviour

and

unlawful coordination.

The US FTC and DOJ have specifically addressed algorithmic pricing in the hotel sector, stating that companies cannot use an algorithm to accomplish conduct that would be unlawful if carried out by humans.

Thus, "the algorithm did it" cannot by itself constitute a competition-law defence.

7. AI as a Gatekeeper

A hyper-intelligent platform may become the intermediary through which consumers access:

• information;
• products;
• financial services;
• software;
• transportation;
• healthcare; and
• entertainment.

The platform's AI assistant may effectively determine:

what the consumer sees → what the consumer considers → what the consumer buys.

This creates a new form of market power:

Cognitive intermediation power

The platform does not merely control access to consumers. It may influence the consumer's decision-making pathway.

8. Six Important Case Laws

Case 1: Google Search (Shopping) — European Union

Case: Google and Alphabet v European Commission, Case T-612/17.

Principle

Google's dominance in general search was connected to the treatment of its specialised comparison-shopping service.

The Court examined whether preferential positioning of Google's own service constituted abusive conduct.

Relevance to hyper-intelligent economies

The case establishes an important principle for AI platforms:

A dominant intermediary cannot necessarily use control over an important gateway to systematically favour its own downstream service.

An AI assistant that recommends its own products instead of objectively comparable rival products could therefore raise analogous competition concerns.

The UK government has also identified Google Shopping as an important example of algorithmic self-preferencing.

Case 2: Google Android — European Union

Case: Google and Alphabet v Commission, Case T-604/18.

The case concerned Google's Android ecosystem and practices involving search, browsers and mobile applications.

Competition significance

The case demonstrates how competition problems can arise where a powerful platform controls multiple complementary technological layers.

Hyper-intelligent-economy application

A comparable AI ecosystem could involve:

operating system + cloud + AI model + assistant + app store + search + advertising.

If competitors must operate through the dominant ecosystem, contractual or technical restrictions may become exclusionary.

Case 3: Google AdSense — European Union

Case: Google and Alphabet v Commission, Case T-334/19.

The dispute concerned contractual restrictions in online search advertising.

Principle

Restrictions imposed by a dominant undertaking on intermediaries can become problematic when they prevent competing providers from obtaining meaningful market access.

AI relevance

An AI platform could potentially impose restrictions through:

• API terms;
• model-access conditions;
• cloud contracts;
• developer restrictions;
• data-access conditions; or
• distribution agreements.

Thus, contractual architecture can become a competition-law instrument.

Case 4: Epic Games v Apple — United States

Case: Epic Games, Inc. v Apple Inc.

This litigation concerned Apple's App Store ecosystem, including distribution restrictions, payment mechanisms and Apple's control over access to iOS consumers.

Hyper-intelligent-economy relevance

AI ecosystems may develop similar bottlenecks.

For example:

AI assistant → app ecosystem → payment system → consumer.

If an AI platform becomes the principal interface through which consumers discover and purchase digital services, control over that interface may become analogous to control over an app-distribution ecosystem.

The competition question becomes whether developers can realistically reach consumers without using the dominant AI intermediary.

Case 5: United States v Google — Search Monopoly Litigation

The US search-monopoly litigation concerning Google provides another important example of competition problems surrounding control over a major digital gateway.

Competition significance

Search engines are not simply ordinary products. They influence:

• information access;
• advertising;
• consumer discovery;
• traffic distribution; and
• downstream commercial opportunities.

Hyper-intelligent-economy relevance

Generative AI could replace conventional search as the primary gateway to information.

An AI assistant could therefore become:

the search engine + recommender + shopping intermediary + advertising interface.

Competition law may consequently need to examine whether AI-generated answers systematically divert opportunities toward the platform's own products.

Case 6: Cornish-Adebiyi v Caesars Entertainment — Algorithmic Pricing

Case: Cornish-Adebiyi v Caesars Entertainment.

The case concerns allegations involving hotel-room pricing and the use of algorithmic pricing systems.

The FTC and DOJ filed a statement of interest emphasizing that an algorithm cannot be used to accomplish conduct that would be unlawful if carried out directly by competitors.

Principle

An algorithm can be:

• the instrument of coordination;
• a mechanism facilitating information exchange; or
• a technological means of implementing an anti-competitive agreement.

Hyper-intelligent relevance

The problem becomes considerably more complicated where AI systems:

• communicate indirectly;
• learn from market signals;
• predict competitors' responses; and
• autonomously change prices.

Competition law therefore needs to distinguish legitimate autonomous optimisation from unlawful coordination.

9. Additional Important Authorities

Several other digital-competition matters provide useful principles for hyper-intelligent markets.

7. Microsoft — Internet Explorer

The Microsoft proceedings demonstrate how tying and leveraging can become important where a dominant firm controls a technological platform and integrates complementary services.

8. Microsoft — Activision Blizzard

The transaction illustrates modern merger-control concerns involving:

• ecosystems;
• gaming platforms;
• cloud services;
• content;
• distribution;
• vertical integration.

9. Amazon Marketplace investigations

Amazon's marketplace model raises questions concerning the use of marketplace data and possible advantages enjoyed by a platform that simultaneously operates as intermediary and seller.

10. Qualcomm

The Qualcomm litigation demonstrates the importance of technology licensing, interoperability and control over technological standards.

These principles can become increasingly significant where AI systems depend upon proprietary technological standards and interfaces.

10. AI Interoperability as a Competition Remedy

Interoperability may become one of the most important tools of competition governance.

Suppose Platform A controls:

• an operating system;
• a large user base;
• an AI assistant;
• search data.

If rival AI assistants cannot access the necessary operating-system functionality, competition may be weakened.

The European Commission has already adopted DMA measures concerning AI interoperability on Android and access by competing search engines to Google Search data. The stated objective is to allow competing AI services to obtain equal access to relevant Android functionality and improve competitive access to search data.

This represents a significant movement toward technical competition governance.

11. AI Data Access

Data access may become analogous to access to essential infrastructure in certain circumstances.

Possible remedies include:

A. Data portability

Consumers can move their data.

B. Data interoperability

Different systems can communicate.

C. Non-discriminatory access

Dominant firms cannot arbitrarily discriminate between competitors.

D. Privacy-preserving data sharing

Data can be shared through:

• anonymisation;
• secure computation;
• federated learning; and
• privacy-enhancing technologies.

Competition law must nevertheless avoid converting every valuable dataset into a mandatory-access resource.

12. AI Mergers and Acquisitions

Traditional merger thresholds may fail to capture important AI transactions.

A company may acquire:

• a small AI startup;
• its engineers;
• its model;
• its dataset;
• intellectual property; or
• technological capabilities.

The target may have low revenue but enormous future competitive significance.

This creates concern about:

"Acqui-hiring"

and

"Killer acquisitions."

Competition authorities therefore increasingly need to consider:

• innovation competition;
• potential competition;
• access to data;
• computing resources;
• talent;
• ecosystem effects; and
• future technological trajectories.

13. Vertical Integration in Hyper-Intelligent Markets

A single company could potentially operate across:

Semiconductors → cloud → foundation model → operating system → AI assistant → marketplace → payment → advertising.

This creates powerful vertical integration.

Vertical integration is not automatically unlawful.

The competition question is whether control at one level enables the undertaking to:

1. foreclose rivals;
2. discriminate against competitors;
3. increase switching costs;
4. restrict interoperability;
5. leverage dominance into adjacent markets; or
6. exploit sensitive competitor information.

14. AI and Predatory Pricing

AI makes dynamic pricing extremely sophisticated.

A dominant company could potentially use:

• personalised pricing;
• real-time discounts;
• predictive demand models;
• cross-subsidisation;
• automated promotional strategies.

Competition authorities may therefore need to examine not merely:

"What is today's price?"

but:

"How does the algorithm determine the price, and what competitive objective does the system pursue?"

15. Personalised Pricing and Consumer Exploitation

Hyper-intelligent systems can predict:

• willingness to pay;
• urgency;
• purchasing history;
• behavioural preferences;
• switching probability.

This could facilitate highly individualised pricing.

Competition law may intersect with:

• consumer protection;
• privacy law;
• data protection;
• discrimination law.

The same dataset may therefore have both competition value and regulatory significance.

16. Autonomous AI Agents

A major future development is the AI purchasing agent.

Instead of a consumer manually comparing:

Product A vs Product B vs Product C,

an AI agent may independently:

1. search the market;
2. compare prices;
3. negotiate;
4. select suppliers;
5. make payment;
6. arrange delivery.

This creates a new intermediary between consumers and sellers.

Competition concerns include:

• discriminatory recommendations;
• exclusion of suppliers;
• preferential treatment;
• commission structures;
• ranking manipulation;
• collusion between agents; and
• control over transaction data.

17. Algorithmic Transparency

Competition authorities may increasingly require firms to explain:

• ranking criteria;
• pricing logic;
• recommendation systems;
• access conditions;
• model dependencies;
• data sources; and
• changes to algorithms.

However, complete disclosure of source code may not always be necessary.

A proportionate approach could involve:

Tier 1 — outcome monitoring

Examine observable market effects.

Tier 2 — technical auditing

Examine algorithmic behaviour.

Tier 3 — controlled regulatory access

Allow authorities to examine confidential technical material.

Tier 4 — remedial intervention

Require modification where competition is materially harmed.

18. Competition Governance Model

A suitable governance framework for hyper-intelligent economies can be represented as:

AI Infrastructure

↓

Data + Compute + Models

↓

Platforms and Ecosystems

↓

AI Intermediaries

↓

Consumers and Businesses

↓

Market Outcomes

Competition governance should operate at every level.

19. Ex-Ante and Ex-Post Regulation

Ex-post competition law

Acts after potentially harmful conduct occurs.

Examples:

• cartel investigation;
• abuse-of-dominance proceedings;
• merger enforcement.

Ex-ante regulation

Establishes obligations before harm occurs.

Examples:

• interoperability;
• data portability;
• non-discrimination;
• transparency;
• access requirements.

The EU's DMA demonstrates this shift. The Commission has used the DMA not merely to punish historical conduct but also to impose forward-looking obligations concerning interoperability, data access and self-preferencing.

20. Competition Governance and Regulatory Sandboxes

AI competition regulation can benefit from controlled regulatory experimentation.

A competition sandbox could allow regulators to examine:

• algorithmic pricing;
• AI agents;
• automated procurement;
• recommendation systems;
• data-sharing arrangements.

Companies could test innovative systems while regulators monitor competition effects.

21. Institutional Requirements

Competition authorities will increasingly require expertise in:

Legal analysis

• antitrust law;
• merger control;
• intellectual property;
• data protection.

Economics

• market definition;
• network effects;
• switching costs;
• innovation effects.

Technology

• AI architecture;
• machine learning;
• cloud infrastructure;
• APIs;
• model training.

Forensics

• algorithmic auditing;
• source-code analysis;
• data-flow analysis;
• digital evidence.

The future competition authority is therefore likely to be partly a legal institution and partly a technical institution.

22. Key Competition Concerns

23. Essential Facilities and AI

The traditional essential-facilities doctrine may become relevant where a technological resource is indispensable for competition.

Possible candidates could theoretically include:

• particular datasets;
• interoperability interfaces;
• infrastructure;
• technical standards;
• computing resources.

But not every valuable AI resource should automatically qualify as an essential facility.

Authorities would ordinarily need to examine:

1. indispensability;
2. absence of realistic alternatives;
3. dominance;
4. competitive foreclosure;
5. objective justification; and
6. proportionality of access.

24. Intellectual Property and Competition

AI markets heavily depend upon intellectual property.

Potential conflicts may involve:

• patents;
• copyrights;
• trade secrets;
• model weights;
• datasets;
• standards-essential patents.

Competition law must therefore maintain a balance between:

Innovation incentives

and

prevention of exclusionary conduct.

Excessively aggressive compulsory access can reduce incentives to innovate, whereas excessive control can potentially entrench dominance.

25. Competition and AI Safety

AI companies may need to cooperate concerning:

• safety testing;
• security;
• incident reporting;
• model evaluation;
• technical standards.

Such cooperation can generate legitimate public benefits.

But cooperation concerning safety cannot automatically immunise otherwise unlawful coordination.

The competition question is:

Is the cooperation genuinely necessary and proportionate to achieving the legitimate safety objective?

This becomes particularly important as AI companies increasingly cooperate on technical standards and safety frameworks.

26. New Theory of Harm: Cognitive Foreclosure

One of the most significant future concepts may be cognitive foreclosure.

Traditional foreclosure prevents competitors from accessing:

inputs, suppliers, distributors or customers.

Cognitive foreclosure could prevent competitors from being considered by consumers at all.

For example:

Consumer asks AI assistant

↓

AI chooses which products to mention

↓

Consumer sees only two recommended products

↓

Other competitors receive no meaningful exposure

↓

Consumer never evaluates alternatives.

Thus, market power may arise from controlling consumer attention and consideration, rather than merely controlling distribution.

27. New Theory of Harm: Algorithmic Dependency

A competitor may remain technically independent while becoming commercially dependent upon a dominant AI infrastructure provider.

For example:

Startup

↓ depends on

Cloud

↓ depends on

Foundation model

↓ depends on

AI API

↓ depends on

dominant platform

Such dependencies may create substantial barriers to entry.

Competition authorities therefore need to examine the entire technological stack rather than analysing every market in isolation.

28. Remedies

Possible remedies include:

Behavioural remedies

• non-discrimination;
• transparency;
• contractual restrictions;
• algorithmic monitoring.

Technical remedies

• interoperability;
• API access;
• data portability;
• compatibility requirements.

Structural remedies

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

Merger remedies

• licensing;
• access commitments;
• data separation;
• interoperability commitments.

Governance remedies

• independent compliance monitors;
• algorithmic audits;
• periodic reporting;
• regulatory access.

Recent Google proceedings illustrate how competition remedies can increasingly include operational and compliance requirements rather than simply monetary penalties. In September 2026, a US federal court required significant changes to Google's advertising practices and imposed monitoring-related measures following findings concerning its ad-tech monopoly; Google has indicated that it will appeal parts of the liability decision.

29. Challenges for Competition Authorities

A. Speed

AI markets can change faster than traditional investigations.

B. Technical complexity

Authorities may struggle to understand proprietary models.

C. Market definition

Traditional product markets may become unstable.

D. Innovation

Intervention may accidentally discourage legitimate innovation.

E. Evidence

Algorithmic decisions may not produce ordinary documentary evidence.

F. Explainability

Machine-learning systems may be difficult to interpret.

G. Globalisation

AI markets operate across jurisdictions.

30. Proposed Hyper-Intelligent Competition Governance Framework

A comprehensive model can be formulated as:

1. Identify

Identify AI-dependent markets and critical infrastructure.

2. Map

Map:

• data flows;
• model dependencies;
• APIs;
• cloud dependencies;
• distribution channels.

3. Measure

Measure:

• market share;
• switching costs;
• network effects;
• data advantages;
• compute access;
• entry barriers.

4. Audit

Audit important algorithmic systems.

5. Interoperate

Promote technically feasible interoperability.

6. Monitor

Continuously monitor dominant AI ecosystems.

7. Investigate

Apply traditional competition law to:

• collusion;
• exclusion;
• tying;
• discrimination;
• abuse of dominance.

8. Remedy

Use proportionate behavioural, technical or structural remedies.

9. Reassess

Continuously reassess markets because technological conditions change rapidly.

31. Key Doctrinal Principles

The following principles are particularly important:

Principle 1 — Algorithmic neutrality

Competition law applies irrespective of whether conduct is performed by a human or an algorithm.

Principle 2 — Technological neutrality

The law should regulate competitive effects rather than favouring or penalising particular technologies.

Principle 3 — Data non-discrimination

Dominant platforms should not arbitrarily discriminate in access to strategically important data where competition law requires access.

Principle 4 — Interoperability

Interoperability can prevent ecosystem foreclosure.

Principle 5 — Algorithmic accountability

Autonomous decision-making does not eliminate corporate responsibility.

Principle 6 — Innovation protection

Competition enforcement must preserve legitimate incentives to innovate.

Principle 7 — Dynamic regulation

Governance must evolve as AI capabilities and market structures change.

32. Conclusion

Competition law in a hyper-intelligent economy must evolve from regulating merely human commercial behaviour to governing technologically mediated market power.

The principal risks include:

• concentration of data;
• concentration of computing power;
• foundation-model dominance;
• algorithmic collusion;
• self-preferencing;
• AI-mediated exclusion;
• ecosystem lock-in;
• discriminatory interoperability;
• killer acquisitions;
• control over AI agents; and
• cognitive foreclosure.

The traditional doctrines of cartels, abuse of dominance, tying, refusal to deal, essential facilities, vertical restraints and merger control remain relevant, but they must be applied to substantially more complex technological environments.

The emerging regulatory architecture therefore combines:

Competition law + AI governance + data governance + interoperability + algorithmic accountability + merger control.

The Google Shopping litigation demonstrates the continuing importance of algorithmic self-preferencing; the algorithmic-pricing litigation represented by Cornish-Adebiyi v Caesars illustrates the difficulties created when pricing decisions are delegated to algorithms; and contemporary DMA measures concerning AI interoperability and search-data access demonstrate the movement toward ex-ante competition governance.