Competition Law and Future-Oriented Antitrust Governance for Computational Societies

1. Introduction

A computational society is an economic and social environment in which increasingly important decisions are made, assisted, or optimized by algorithms, artificial intelligence, automated pricing systems, recommendation engines, data infrastructures, cloud computing, digital platforms, and machine-learning models.

Traditional competition law was largely designed around identifiable firms, observable prices, conventional markets, and human decision-makers. Computational markets create a different environment:

• prices can change automatically within seconds;
• competitors can use the same pricing algorithm;
• algorithms can learn from competitors' conduct;
• platforms can rank their own products;
• access to data can become a competitive bottleneck;
• AI models can become essential inputs;
• switching and interoperability can be technically controlled;
• acquisitions of emerging technologies may occur before conventional market power becomes visible;
• automated systems can produce exclusionary effects without a traditional written agreement.

Consequently, future-oriented antitrust governance means adapting competition law so that it can identify and remedy competitive harm produced by computational systems while preserving innovation and legitimate technological competition.

Recent enforcement demonstrates that this is no longer purely theoretical. The EU's Digital Markets Act has moved toward ex ante regulation of major digital gatekeepers, while U.S. enforcement has increasingly addressed algorithmic pricing and data-driven platform conduct.

2. Meaning of Computational Society

A computational society has several defining characteristics.

A. Algorithmic decision-making

Businesses increasingly delegate:

• pricing;
• inventory allocation;
• advertising;
• credit decisions;
• search rankings;
• product recommendations;
• resource allocation;
• fraud detection;
• customer targeting

to algorithms.

The competitive problem arises when an algorithm does not merely improve efficiency but coordinates, excludes, discriminates, or strategically disadvantages competitors.

B. Data as a competitive resource

Large-scale data can provide advantages in:

• prediction;
• personalization;
• AI training;
• advertising;
• search;
• logistics;
• financial services;
• healthcare;
• consumer profiling.

The European Commission's 2026 Google Search-data proceedings illustrate the emerging regulatory importance of access to competitively significant data. The Commission required measures concerning sharing of anonymised Google Search data with eligible competitors, including AI search services.

C. Platform ecosystems

A computational firm may simultaneously operate:

operating system → app store → payment system → advertising → cloud → AI → data infrastructure.

Competition therefore may occur between ecosystems, rather than merely between individual products.

D. Machine-speed competition

Human regulators operate at a fundamentally slower speed than computational markets.

An algorithm can:

1. observe competitors;
2. alter its strategy;
3. test thousands of price combinations;
4. learn from consumer responses; and
5. implement the resulting strategy

within seconds.

This creates difficulties for conventional investigations.

3. Traditional Antitrust Versus Computational Antitrust

The transformation does not mean traditional antitrust principles disappear. Rather, the factual mechanisms through which those principles operate change.

4. Core Objectives of Future-Oriented Antitrust Governance

A future-oriented system should address at least eight objectives.

4.1 Prevent algorithmic collusion

Competition authorities must distinguish:

• legitimate parallel conduct;
• conscious adaptation to market conditions;
• unilateral algorithmic learning;
• tacit coordination;
• explicit human coordination implemented through algorithms.

The central issue is whether computational technology merely facilitates competition or becomes an instrument of coordination.

4.2 Control algorithmic self-preferencing

A vertically integrated platform may control:

• the search algorithm;
• ranking system;
• marketplace;
• payment infrastructure;
• advertising network.

It can therefore potentially design its algorithm to advantage its own downstream products.

The Google Shopping litigation is particularly important. In 2024, the Court of Justice upheld the finding that Google abused its dominant position by favouring its own comparison-shopping service over competing services.

This demonstrates that algorithmic ranking can constitute a competition-law mechanism, rather than merely a technical design choice.

5. Case Laws

Case 1 — Eturas UAB and Others v Lithuanian Competition Council, C-74/14

Facts

Travel agencies used a common computerized booking system. The system administrator sent a message concerning a limitation on online discounts, and the system automatically restricted discounts available to customers.

Legal significance

The CJEU considered whether participation in such a computerized environment could constitute a concerted practice under Article 101 TFEU.

The Court emphasized issues concerning knowledge, participation, evidence and the possibility of rebutting an inference of participation.

Computational significance

Eturas is foundational because it shows that competition law can operate where:

technology mediates the mechanism through which competitors' conduct becomes coordinated.

It is particularly relevant to future platforms using common software architectures.

Case 2 — United States v David Topkins

Facts

David Topkins and co-conspirators sold posters through Amazon Marketplace. The participants agreed to fix prices and used algorithmic repricing software to implement the arrangement.

The DOJ described it as its first criminal prosecution specifically targeting an e-commerce price-fixing conspiracy.

Legal significance

The case demonstrates that:

the use of an algorithm does not immunize an otherwise conventional price-fixing agreement from antitrust liability.

Future lesson

Authorities should investigate the human arrangement behind the algorithm, rather than treating automated execution as an independent technological phenomenon.

Case 3 — Trod Ltd and GB eye Ltd

The UK's Competition and Markets Authority investigated two competing sellers on Amazon Marketplace that agreed not to undercut one another.

They used automated repricing software to implement their agreement. The CMA imposed a fine exceeding £160,000.

Significance

This case demonstrates a crucial principle:

An algorithm can be the implementation mechanism for a cartel even when humans make the underlying agreement.

Therefore, future antitrust governance must examine:

• software settings;
• pricing rules;
• API communications;
• algorithmic instructions;
• synchronization mechanisms;
• historical price data.

Case 4 — United States v Apple Inc. — E-Books

Facts

The U.S. government successfully challenged Apple's arrangements with major publishers concerning e-book pricing.

The U.S. District Court found Apple had conspired to fix e-book prices. The subsequent remedy included termination of relevant agreements and measures intended to restore competition.

Computational significance

The case predates today's generative-AI environment, but it illustrates an important principle for computational societies:

technological platforms can alter the architecture of competition even when the underlying antitrust problem remains conventional price coordination.

Future cases may involve:

• AI marketplaces;
• model marketplaces;
• app stores;
• cloud platforms;
• digital-content distribution;
• AI-generated content platforms.

Case 5 — Google LLC and Alphabet Inc. v European Commission — Google Shopping, C-48/22 P

Facts

Google operated a general search service while also operating its own comparison-shopping service.

The European Commission found that Google favoured its own comparison-shopping results over competing services.

In September 2024, the CJEU dismissed Google's appeal and upheld the €2.4 billion fine.

Legal significance

The case is exceptionally important for computational competition because the alleged exclusionary mechanism operated through:

• ranking;
• visibility;
• search algorithms;
• traffic allocation;
• platform architecture.

The Court addressed issues including foreclosure capability, causal connection and competition on the merits.

Future lesson

Ranking systems can function as competitive infrastructure.

Consequently, antitrust governance should examine not merely prices but also:

who receives computational visibility?

Case 6 — Google Search and Search Advertising — United States

The U.S. government's case against Google concerned Google's maintenance of monopolies in general search and search advertising through exclusionary distribution arrangements.

In August 2024, the District Court found Google liable under Section 2 of the Sherman Act. In April 2026, the DOJ announced remedies requiring, among other things, certain competitors to receive access to search-index and user-interaction data and search/text-ad syndication services.

Computational significance

This case illustrates a transition from:

market power → exclusionary contracts

toward a broader concern:

market power → control of computational infrastructure → reinforcement of market power.

Search data can improve algorithms, which can increase quality, which can attract users, which can generate more data.

This produces a potential:

data → algorithmic quality → users → more data

feedback loop.

Future antitrust governance therefore has to examine data-feedback effects, not merely static market share.

Case 7 — United States and States v RealPage, Inc.

Facts

The DOJ brought an antitrust action against RealPage concerning algorithmic rental pricing.

According to the DOJ's allegations, competing landlords supplied non-public, competitively sensitive information concerning rents, vacancies and lease terms to RealPage's software, which then generated pricing recommendations. The case alleges violations of Sections 1 and 2 of the Sherman Act.

Significance

RealPage is particularly important because the alleged competitive mechanism is not simply:

competitors communicate directly.

Instead:

competitors provide information → common computational system processes information → pricing recommendations → competing firms implement recommendations.

This raises the question whether an algorithmic intermediary can become a coordination infrastructure.

The DOJ's 2026 proceedings also show continuing development of remedies addressing algorithmic pricing systems.

Case 8 — Apple and Meta under the EU Digital Markets Act

In April 2025, the European Commission found Apple in breach of the DMA's anti-steering obligation and Meta in breach of an obligation concerning consumer choice regarding use of personal data. The Commission imposed fines of €500 million on Apple and €200 million on Meta.

Significance

This represents a movement from purely ex-post antitrust toward ex-ante digital competition governance.

Instead of waiting for a lengthy Article 102 investigation to establish all elements of abuse, the DMA imposes specific obligations on designated gatekeepers.

This is highly relevant to computational societies because some digital advantages can become entrenched rapidly.

6. Major Future Competition Problems

A. Algorithmic Collusion

Algorithms may facilitate coordination through:

• common pricing software;
• common data providers;
• predictive pricing;
• reinforcement learning;
• automated retaliation;
• market-monitoring systems.

The critical question becomes:

When does autonomous adaptation become legally relevant coordination?

Competition authorities may need to examine the entire architecture rather than searching only for traditional communications.

B. Algorithmic Tacit Coordination

Two algorithms can potentially observe market prices and adjust their own prices accordingly.

Even without direct communication, repeated interaction may generate stable pricing patterns.

The legal challenge is distinguishing:

Legitimate behaviour

“Prices are changing because market conditions changed.”

from

Potentially problematic behaviour

“The system is deliberately designed to recognize competitors' behaviour and maintain supra-competitive coordination.”

7. Data Concentration and Competition

Data can create several forms of market power.

1. Data scale advantage

More users → more data → better model → better service → more users.

2. Data exclusivity

A dominant firm may deny rivals access to data necessary to compete.

3. Data combination

A company may combine information from:

• search;
• social media;
• payments;
• advertising;
• e-commerce;
• location;
• cloud services.

4. Data portability barriers

Users may technically own or access their data but face practical barriers to transferring it.

Thus future antitrust analysis may have to distinguish:

data ownership, data access, data portability, data interoperability and data-derived competitive advantage.

8. AI and Computational Competition

Generative AI introduces another layer.

Competition may depend on access to:

• GPUs;
• cloud computing;
• training datasets;
• foundation models;
• model weights;
• inference infrastructure;
• AI talent;
• distribution platforms;
• application programming interfaces.

A future AI ecosystem could therefore contain several vertically related layers:

chips → cloud → foundation model → API → application → distribution → consumer data

Control over one layer may reinforce market power in another.

The European Commission's 2026 cloud proceedings illustrate the direction of this development: the Commission stated that AWS and Azure were provisionally considered candidates for DMA gatekeeper designation in cloud services, citing entrenched positions, switching costs, ecosystems and AI-related factors.

9. Cloud Computing as Antitrust Infrastructure

Cloud services increasingly constitute foundational infrastructure for:

• AI;
• fintech;
• e-commerce;
• government services;
• software;
• healthcare;
• digital media.

Competition concerns can therefore involve:

• switching costs;
• interoperability;
• data portability;
• egress fees;
• bundling;
• preferential treatment;
• technical restrictions;
• contractual lock-in.

The European Commission was still conducting a DMA market investigation into cloud computing in 2026, including interoperability and contractual conditions.

10. Computational Gatekeepers

Future competition law should recognize a category of firms that function as computational gatekeepers.

A computational gatekeeper may control:

• search;
• ranking;
• identity;
• payments;
• app distribution;
• cloud infrastructure;
• advertising;
• AI models;
• data access.

Its power may arise less from a conventional monopoly over a product and more from its ability to control the rules by which other businesses reach consumers.

The EU DMA represents one institutional response: Alphabet, Amazon, Apple, ByteDance, Meta and Microsoft were initially designated as gatekeepers in 2023.

11. Algorithmic Self-Preferencing

Self-preferencing can take multiple computational forms:

1. ranking one's own products higher;
2. allocating better search visibility;
3. giving one's services preferential recommendation;
4. prioritizing one's own advertising;
5. restricting rivals' data access;
6. changing API functionality;
7. designing default settings.

The Google Shopping litigation demonstrates why algorithmic neutrality and ranking architecture can become central antitrust questions.

12. Predictive and Personalized Pricing

Traditional price discrimination generally involves different prices for different consumers.

Computational pricing can be considerably more sophisticated.

Algorithms may use:

• browsing history;
• location;
• purchasing history;
• device information;
• time;
• predicted willingness to pay;
• competitor prices;
• inventory;
• consumer segmentation.

Competition law must therefore investigate whether personalization:

• improves efficiency;
• reflects legitimate cost differences;
• exploits market power;
• facilitates coordination;
• excludes competitors.

13. Computational Mergers

Traditional merger review asks whether:

Firm A + Firm B = excessive concentration.

Computational merger review may additionally need to consider:

data + algorithms + talent + cloud infrastructure + distribution = future market power.

A small AI company may have:

• relatively little revenue;
• limited current market share;
• but highly valuable technology or data.

Consequently, traditional turnover thresholds may fail to identify strategically important acquisitions.

Future merger governance may therefore consider:

• data assets;
• AI models;
• intellectual property;
• developer ecosystems;
• computational capacity;
• user networks;
• potential competition.

14. Killer Acquisitions in Computational Markets

A dominant platform may acquire a small technological firm before the latter becomes a meaningful competitor.

The competitive concern is not necessarily current market share.

It may instead be:

What competitive trajectory has been eliminated?

This requires greater emphasis on:

• innovation competition;
• pipeline products;
• technological substitutes;
• developer communities;
• nascent platforms;
• potential entrants.

15. Interoperability as an Antitrust Remedy

Interoperability can be used to reduce ecosystem lock-in.

Possible remedies include:

• API access;
• data portability;
• protocol interoperability;
• messaging interoperability;
• payment interoperability;
• cloud interoperability;
• search-data access.

The objective is not necessarily to make every system identical.

Instead:

interoperability reduces the ability of a dominant technical architecture to become an unavoidable competitive bottleneck.

16. Algorithmic Auditing

Future competition authorities may need technical audit capabilities.

An algorithmic investigation could examine:

Input layer

• What data enters the system?

Processing layer

• What variables influence the decision?

Model layer

• What optimization objective is used?

Output layer

• What decision does the system generate?

Feedback layer

• Does the output influence the next round of learning?

Governance layer

• Who can modify the model?

This creates a new form of competition-law evidence.

17. Computational Evidence

Future investigations may increasingly depend on:

• source code;
• model cards;
• audit logs;
• API logs;
• version histories;
• datasets;
• training records;
• server records;
• pricing histories;
• experiment records;
• internal communications;
• model outputs.

The evidentiary question becomes:

How should a competition authority prove an anticompetitive computational mechanism when the mechanism is probabilistic, adaptive or continuously changing?

18. Explainability and Due Process

Competition authorities cannot simply say:

“The algorithm produced an anticompetitive result.”

They may need to establish:

1. what the algorithm was designed to do;
2. what inputs it received;
3. how it behaved;
4. whether the outcome was foreseeable;
5. whether human decision-makers influenced it;
6. whether the firm monitored the result;
7. whether alternative competitive designs existed.

This becomes particularly important where enforcement consequences are severe.

19. Remedies for Computational Antitrust Violations

Traditional remedies such as fines may be insufficient where the competitive harm results from architecture.

Future remedies may include:

Structural remedies

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

Behavioural remedies

• non-discrimination;
• interoperability;
• data access;
• anti-steering.

Technical remedies

• API access;
• algorithmic auditing;
• independent monitoring;
• transparency requirements.

Governance remedies

• compliance officers;
• algorithm review boards;
• periodic regulatory reporting;
• independent technical audits.

Data remedies

• portability;
• data-sharing obligations;
• restrictions on combining datasets.

20. Ex-Ante and Ex-Post Governance

A major future policy question is the balance between:

Ex-post antitrust

Act after harmful conduct has occurred.

Examples:

• Article 101/102 TFEU;
• Sherman Act;
• traditional merger enforcement.

Ex-ante digital regulation

Impose obligations before competitive harm becomes entrenched.

The DMA represents this model. The EU has continued developing it in 2026, including proceedings concerning search-data access and cloud services.

A computational society may require both systems simultaneously.

21. Future Regulatory Architecture

A comprehensive future-oriented framework could be represented as:

Computational Market

↓

Identify Digital/Computational Infrastructure

↓

Define Relevant Market/Ecosystem

↓

Measure Market Power

↓

Assess Data + Compute + Network Effects

↓

Audit Algorithmic Conduct

↓

Identify Coordination / Exclusion / Self-Preferencing

↓

Assess Effects on Competition

↓

Consider Innovation and Efficiency

↓

Select Remedy

↓

Continuous Monitoring

This represents a shift from:

one-time competition assessment

to:

continuous competition governance.

22. Special Problems for Competition Authorities

22.1 Regulatory lag

Technology may develop faster than legislation.

22.2 Technical expertise

Traditional lawyers and economists may need assistance from:

• computer scientists;
• data scientists;
• AI researchers;
• cybersecurity specialists.

22.3 Black-box systems

Authorities may not know precisely why a machine-learning model generated an output.

22.4 Cross-border enforcement

A computational platform may operate globally while its competitive effects occur locally.

22.5 Innovation trade-offs

Overregulation can potentially discourage legitimate experimentation.

Therefore, future governance must distinguish:

technology-enabled efficiency from technology-enabled exclusion.

23. Important Principles for Future Computational Antitrust

Principle 1 — Technology neutrality

The fact that conduct is automated should neither create immunity nor automatically establish liability.

Principle 2 — Substance over software

Authorities should examine the economic effect of the computational system.

Principle 3 — Human responsibility

Companies should not escape responsibility merely because a machine performed the final action.

Principle 4 — Data competition

Data access and data accumulation should form part of competitive analysis where economically relevant.

Principle 5 — Ecosystem analysis

Competition authorities should examine interconnected markets rather than isolated products.

Principle 6 — Innovation sensitivity

Remedies should preserve legitimate innovation where possible.

Principle 7 — Continuous supervision

Highly dynamic computational markets may require continuing monitoring rather than a single enforcement event.

24. Synthesis of the Case Laws

25. Emerging Doctrine: From Market Power to Computational Power

A particularly important future development is the expansion of the concept of market power.

Traditional analysis asks:

What share of the relevant market does the firm possess?

Computational analysis may additionally ask:

What computational capabilities does the firm control?

These may include:

• data;
• compute;
• AI models;
• APIs;
• algorithms;
• digital identity;
• cloud infrastructure;
• distribution;
• interoperability standards;
• user attention.

Accordingly, future antitrust analysis may increasingly examine:

Market power = economic position + data advantage + computational capacity + network effects + ecosystem control + switching costs.

This is not a replacement for established legal tests, but a framework for identifying where conventional tests may need technically informed application.

26. Conclusion

Future-oriented antitrust governance for computational societies represents the transition from competition law focused primarily on firms, prices and conventional agreements toward competition governance capable of addressing algorithms, data, AI, cloud infrastructure, platforms, ecosystems and machine-mediated coordination.

The existing cases already provide the building blocks:

• Eturas demonstrates the relevance of computer-mediated coordination;
• Topkins and Trod/GB eye demonstrate algorithmic implementation of price fixing;
• Apple E-books demonstrates digital-platform facilitation of conventional coordination;
• Google Shopping establishes the importance of algorithmic ranking and self-preferencing;
• the U.S. Google litigation illustrates the significance of distribution and data infrastructure;
• RealPage brings algorithmic pricing directly into contemporary antitrust enforcement;
• the DMA demonstrates movement toward ex-ante governance of computational gatekeepers.

The central challenge for the future is therefore not simply “regulating algorithms.” It is designing a competition regime capable of answering three interconnected questions:

Who controls the computational infrastructure?

How does that infrastructure influence competitive behaviour?

What regulatory intervention preserves competition without unnecessarily suppressing technological innovation?