Competition Law and Future Governance of Machine-Mediated Commerce

1. Introduction

Machine-mediated commerce refers to markets in which commercial decisions are substantially made, recommended, negotiated, executed, or modified by software systems rather than directly by human decision-makers. Examples include:

• AI pricing systems;
• algorithmic purchasing agents;
• autonomous procurement systems;
• recommendation engines;
• automated bidding systems;
• digital marketplaces;
• smart contracts;
• AI-driven supply-chain allocation;
• autonomous advertising systems;
• algorithmic inventory management;
• machine-to-machine transactions; and
• AI agents that negotiate prices and contractual terms.

The central competition-law problem is that traditional antitrust law generally examines human firms making commercial decisions, whereas future markets may involve machines independently making thousands or millions of decisions in real time.

The governing question therefore becomes:

Who is responsible when machines, algorithms, and autonomous commercial agents produce anticompetitive market outcomes?

Future competition governance will have to combine conventional rules on agreements, concerted practices, abuse of dominance and merger control with algorithmic accountability, interoperability, auditability, data governance and technological remedies.

2. Meaning of Machine-Mediated Commerce

Machine-mediated commerce exists where software materially intervenes between suppliers and consumers or between competing businesses.

A simplified model is:

Human/Company → Algorithm/AI Agent → Market Decision → Consumer/Competitor

In more advanced markets:

AI Agent A ↔ AI Agent B ↔ Platform ↔ AI Agent C → Transaction

The machines may:

1. determine prices;
2. select suppliers;
3. allocate customers;
4. rank products;
5. recommend transactions;
6. determine discounts;
7. monitor competitors;
8. automatically react to competitors' conduct;
9. negotiate contracts; and
10. execute transactions without further human approval.

This creates a fundamental competition-law difficulty: the absence of an explicit human agreement does not necessarily mean that competition is functioning normally.

3. Existing Competition-Law Framework

Machine-mediated commerce can presently be analysed through established competition-law concepts.

A. Anti-competitive agreements

Algorithms may facilitate:

• price fixing;
• market allocation;
• customer allocation;
• bid coordination;
• resale-price maintenance;
• information exchange; and
• exclusionary agreements.

B. Abuse of dominance

A dominant digital platform or AI intermediary may use algorithms to:

• self-preference;
• discriminate against rivals;
• restrict interoperability;
• deny access to essential data;
• manipulate rankings;
• impose discriminatory conditions;
• tie complementary services; or
• exclude competing AI agents.

C. Merger control

Competition authorities may increasingly have to examine acquisitions involving:

• AI models;
• training datasets;
• cloud infrastructure;
• algorithmic distribution channels;
• digital marketplaces;
• data brokers; and
• AI-agent ecosystems.

D. Digital-market regulation

Future governance may supplement conventional antitrust with ex ante obligations concerning:

• interoperability;
• data portability;
• access;
• transparency;
• switching;
• algorithmic neutrality; and
• platform conduct.

4. Why Machine-Mediated Commerce Creates New Competition Risks

4.1 Algorithmic Price Coordination

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

The difficult question is whether:

parallel algorithmic conduct can amount to unlawful coordination without an express human agreement.

Traditional competition law distinguishes between:

• independent parallel conduct; and
• coordinated conduct resulting from communication or agreement.

Machine-mediated markets can blur that distinction.

5. Case Law

Case 1 — Eturas UAB v Lietuvos Respublikos Konkurencijos Taryba

Court: Court of Justice of the European Union
Case: C-74/14

Facts

An online travel-booking platform communicated a technical message to participating travel agencies concerning a limitation on discounts that could be offered through the platform.

The system subsequently restricted the maximum discount that could be applied.

Legal significance

The CJEU examined whether businesses could be responsible for participating in an anticompetitive arrangement implemented through an electronic platform.

The Court recognised that participation in a technological system can be relevant to establishing concerted practice where the undertaking knew or could reasonably have been expected to know about the communication and continued participating.

Importance for machine-mediated commerce

The case demonstrates that:

Electronic architecture can become the mechanism through which coordination occurs.

The absence of a traditional face-to-face meeting does not necessarily prevent competition law from applying.

Future relevance

An AI marketplace could similarly communicate:

• recommended pricing;
• discount ceilings;
• inventory restrictions;
• customer-allocation parameters; or
• bidding instructions.

Competition authorities may therefore examine both the algorithm and the firm's knowledge and participation in its operation.

6. Case 2 — United States v Topkins

Court: United States District Court, Northern District of California
Year: 2015

Facts

An online seller used pricing algorithms to coordinate prices for posters sold through an online marketplace.

The case involved an agreement between competitors to adopt pricing algorithms that implemented coordinated pricing.

Competition-law significance

The important principle was that the use of sophisticated software did not change the fundamental nature of the conduct.

An algorithm can be merely the instrument through which an unlawful agreement is implemented.

Relevance

The case establishes an important distinction:

Algorithmic implementation ≠ independent algorithmic conduct.

If humans first agree to coordinate and then instruct software to implement that agreement, conventional cartel law can apply.

7. Case 3 — Competition and Markets Authority v Trod Ltd

Authority: UK Competition and Markets Authority
Year: 2016

Facts

Online sellers agreed to coordinate prices for products sold through an online marketplace.

Pricing software was used to monitor and maintain the agreed pricing arrangements.

Legal significance

The matter demonstrated how automated pricing systems can make an existing cartel more effective.

Instead of employees continuously monitoring competitors, software can:

• observe prices;
• identify deviations;
• adjust prices;
• maintain predetermined relationships.

Importance

Machine systems can therefore increase the speed, frequency and precision of cartel implementation.

Future competition authorities may accordingly investigate:

• source code;
• algorithmic instructions;
• API communications;
• system logs;
• version histories;
• training data; and
• automated decision records.

8. Case 4 — Google Shopping

Case: Commission v Google and Alphabet
EU General Court: T-612/17

Facts

Google operated a general search service and specialised comparison-shopping services.

The European Commission found that Google had given preferential treatment to its own comparison-shopping service in search results.

Competition issue

The case concerned the use of a dominant digital infrastructure to favour an affiliated service.

Importance for machine-mediated commerce

Machine-mediated commerce increasingly depends upon algorithmic ranking.

A platform may determine:

• which seller appears first;
• which product is recommended;
• which supplier receives visibility;
• which AI agent is accessible;
• which advertisement receives attention.

Consequently, ranking algorithms can become a competitive bottleneck.

Future principle

Competition governance may increasingly ask:

Does the operator control the rules by which autonomous commercial agents obtain market visibility?

9. Case 5 — Google Android

Case: Google and Alphabet v European Commission
General Court: T-604/18

Facts

The case concerned Google's conduct involving Android and various contractual restrictions associated with mobile-device manufacturers and distribution.

The EU institutions examined whether Google's contractual arrangements reinforced the dominance of its search service.

Relevance to machine-mediated commerce

The significance extends beyond smartphones.

Future AI ecosystems may involve a similar structure:

Operating system → AI assistant → search → marketplace → payment → advertising

If one undertaking controls several layers of this stack, it may be capable of influencing the commercial decisions made by millions of users or autonomous agents.

Competition concern

The important future question is therefore not merely:

Who sells the product?

It is:

Who controls the technological infrastructure through which commercial decisions are made?

10. Case 6 — United States v Apple Inc. — e-books

Court: U.S. District Court for the Southern District of New York; subsequent appellate proceedings

Facts

Apple was found liable for participating in conduct concerning the pricing of electronic books.

The case involved contractual arrangements with publishers and the restructuring of the distribution model for digital content.

Importance for machine-mediated commerce

The case illustrates the competition risks created when a powerful intermediary changes the architecture through which transactions take place.

In machine-mediated markets, the intermediary may control:

• transaction rules;
• commissions;
• access;
• ranking;
• payment;
• contractual templates; and
• technical compatibility.

Future implication

An AI marketplace that simultaneously controls discovery, negotiation, payment and execution could become a particularly important competition-law gatekeeper.

11. Case 7 — Epic Games v Apple

Court: U.S. District Court for the Northern District of California; Ninth Circuit proceedings

This litigation concerned Apple's control over app distribution and payment mechanisms.

The dispute illustrates competition concerns surrounding platform-controlled commercial ecosystems.

For machine-mediated commerce, the analogy is important because future AI agents may need access to:

• application marketplaces;
• payment systems;
• APIs;
• cloud services;
• identity systems;
• operating systems; and
• app distribution infrastructure.

If access to these systems is controlled by a dominant intermediary, competition questions can arise concerning exclusion, tying, restrictions on alternative payment systems and access to competing distribution channels.

12. Case 8 — Booking.com / HRS and Online Platform Parity

European competition authorities have examined most-favoured-nation/parity clauses used by online platforms.

The central concern is whether a platform can prevent suppliers from offering different prices or conditions through competing channels.

Machine-commerce significance

AI purchasing agents may automatically compare:

• Amazon;
• independent retailers;
• manufacturer websites;
• specialist marketplaces; and
• competing AI marketplaces.

Contractual restrictions preventing suppliers from offering different terms can therefore interfere with the ability of autonomous agents to perform genuine price comparison.

13. Algorithmic Collusion

One of the most important future issues is algorithmic collusion.

There are several possible models.

Model 1 — Human agreement + algorithm

Human competitors agree to fix prices.

The algorithm implements the agreement.

Traditional cartel law applies relatively straightforwardly.

Model 2 — Algorithm + human awareness

No explicit price-fixing agreement exists, but firms deliberately deploy algorithms designed to respond to competitors in a coordinated manner.

The legal analysis becomes more difficult.

Model 3 — Independent algorithmic adaptation

Two algorithms independently learn that maintaining high prices can maximise profits.

Neither company expressly instructed the algorithm to coordinate.

This creates the most difficult problem.

The central question becomes:

Can competition law impose responsibility for an anticompetitive equilibrium produced by autonomous learning?

14. Autonomous AI Agents and Competition

Future consumers may increasingly delegate commercial decisions to AI agents.

For example:

Consumer → AI Agent → compares products → negotiates price → selects supplier → completes transaction.

This changes the competitive structure.

The relevant "consumer" may no longer directly inspect:

• prices;
• advertisements;
• product rankings;
• terms and conditions.

Instead, an AI agent performs these functions.

This creates potential risks of machine-mediated consumer steering.

15. AI Agent Gatekeepers

A dominant AI assistant could influence commerce by determining which suppliers its users see.

For example:

User asks:
"Buy me a laptop under ₹80,000."

The AI agent may select:

• which marketplaces to search;
• which products to display;
• which sellers to exclude;
• which advertisements to consider;
• which payment system to use.

If the AI operator owns a marketplace, cloud service, payment system or advertising network, conflicts of interest may arise.

Competition law will therefore increasingly examine vertical integration across the AI commercial stack.

16. Self-Preferencing

Self-preferencing occurs when a platform gives preferential treatment to its own products or services.

In machine-mediated commerce, self-preferencing could become considerably more sophisticated.

An AI platform might manipulate:

• product rankings;
• recommendations;
• search results;
• default suppliers;
• procurement suggestions;
• automated bidding;
• product descriptions; or
• transaction routing.

The discrimination may be invisible to consumers because the decision is generated by an algorithm.

17. Data as a Competitive Input

Machine-mediated commerce depends heavily upon data.

Important categories include:

• transaction data;
• consumer preferences;
• search histories;
• pricing data;
• supplier performance data;
• logistics information;
• inventory data;
• behavioural data; and
• AI-training datasets.

A dominant platform could potentially use commercially sensitive information obtained from dependent businesses to compete against those businesses.

This raises questions concerning:

data access + data portability + confidentiality + dominance + self-preferencing.

18. Algorithmic Discrimination

Machine-mediated commerce can create discriminatory treatment through automated systems.

Examples include:

• different prices for different consumers;
• different rankings for different sellers;
• discriminatory access to advertising;
• different commission rates;
• exclusion of certain suppliers;
• differential visibility; and
• preferential treatment of affiliated businesses.

Competition authorities will need to distinguish between:

Legitimate differentiation

For example:

• genuine cost differences;
• quality differences;
• risk differences.

and

Anticompetitive discrimination

For example:

• exclusion of rivals;
• discriminatory access;
• strategic foreclosure.

19. Machine-Mediated Vertical Restraints

AI systems can automatically enforce:

• resale-price restrictions;
• territorial restrictions;
• customer restrictions;
• exclusivity;
• MFN clauses;
• bundling;
• tying; and
• loyalty incentives.

The machine therefore transforms a contractual provision into an automatically enforced market rule.

This may make enforcement more effective but can also make violations harder to detect.

20. Interoperability as a Future Competition Remedy

Traditional remedies often involve:

• fines;
• injunctions;
• divestiture;
• contractual modifications.

Machine-mediated markets may require technological remedies.

These could include:

A. API access

Competitors receive access to interfaces necessary to interoperate.

B. Data portability

Users and businesses can transfer relevant data between competing systems.

C. Interoperability

AI agents can communicate with competing platforms.

D. Non-discrimination

The platform must apply equivalent technical conditions to rivals.

E. Algorithmic auditing

Authorities may inspect systems to identify discriminatory or exclusionary conduct.

21. Algorithmic Transparency

Complete disclosure of source code may not always be necessary or desirable.

A future regulatory framework could instead require meaningful algorithmic accountability, including:

• documentation of objectives;
• identification of relevant inputs;
• records of material changes;
• testing for discriminatory effects;
• preservation of decision logs;
• explanation of ranking criteria;
• audit trails; and
• independent compliance assessments.

The objective would be to make automated commercial conduct legally observable without necessarily requiring publication of proprietary source code.

22. Algorithmic Audit Trails

Future competition investigations may increasingly depend on digital evidence.

Important evidence may include:

1. source code;
2. model versions;
3. training data;
4. system prompts;
5. API logs;
6. transaction records;
7. pricing histories;
8. model outputs;
9. human instructions;
10. system configuration;
11. communications between agents; and
12. records of algorithmic modifications.

This produces a new form of competition evidence:

machine-generated evidence of commercial coordination.

23. Responsibility for Autonomous Commercial Decisions

A future governance model could distinguish four levels of responsibility.

Level 1 — Direct human instruction

A company explicitly instructs an AI system to engage in anticompetitive conduct.

Traditional liability principles are relatively straightforward.

Level 2 — Designed functionality

The company intentionally designs an algorithm to produce exclusionary results.

Design responsibility becomes important.

Level 3 — Foreseeable autonomous conduct

The company does not expressly instruct the system but knows that the system repeatedly produces anticompetitive effects.

Monitoring and compliance duties become important.

Level 4 — Unexpected autonomous behaviour

An AI system produces an unexpected market outcome.

The difficult question becomes whether competition law should impose responsibility merely because the undertaking operated the system.

This area will require careful doctrinal development.

24. Autonomous Contracting

AI agents may eventually negotiate contracts without direct human approval.

For example:

AI Buyer ↔ AI Seller

The machines may negotiate:

• price;
• quantity;
• delivery;
• warranties;
• payment;
• exclusivity; and
• renewal.

This creates an important legal question:

When two autonomous agents reach the same commercial outcome, when does their interaction constitute lawful market competition and when does it become coordination?

Competition law will have to remain capable of distinguishing legitimate automated competition from automated collusion.

25. Smart Contracts and Competition Law

Smart contracts can automatically execute transactions when predetermined conditions are satisfied.

Advantages include:

• efficiency;
• lower transaction costs;
• certainty;
• automatic enforcement.

But they can also facilitate:

• automatic price coordination;
• exclusion;
• discriminatory access;
• automatic loyalty arrangements;
• market allocation.

Therefore:

Automation of an agreement does not immunise the agreement from competition law.

26. Machine-Mediated Mergers

Future merger control will increasingly examine acquisitions involving:

• foundation models;
• AI-agent companies;
• data providers;
• cloud platforms;
• marketplaces;
• semiconductor companies;
• robotics companies;
• autonomous logistics systems.

Traditional turnover thresholds may not fully capture the competitive significance of an acquisition involving an emerging AI company with:

• little current revenue;
• enormous data assets;
• strategic technology;
• high switching costs; or
• rapidly expanding network effects.

27. Killer Acquisitions in Machine Commerce

A dominant technology platform might acquire an emerging AI-agent company before it becomes a significant competitor.

Competition authorities may therefore examine:

• innovation competition;
• potential competition;
• access to data;
• interoperability;
• ecosystem effects;
• future market expansion; and
• control of technical standards.

The competitive assessment may need to consider future technological trajectories, rather than merely current market shares.

28. Network Effects

Machine-mediated markets frequently display strong network effects.

More users produce:

more transactions → more data → better algorithms → more users.

This creates a feedback loop:

Users → Data → Intelligence → Better service → More users

Such feedback can generate significant entry barriers.

A competitor may therefore struggle even if it possesses technically comparable software.

29. The Data–AI–Market Power Triangle

A central future competition problem can be represented as:

Data

↓

AI Capability

↓

Market Power

↓

More Transactions

↓

More Data

The resulting cycle can strengthen an incumbent's position.

Competition policy may consequently need to examine not only market share but also:

• data advantages;
• computational resources;
• ecosystem integration;
• switching costs;
• interoperability;
• access to distribution; and
• quality of AI outputs.

30. Future Governance Architecture

A comprehensive governance model could contain six layers.

Layer 1 — Ex-post antitrust

Traditional investigation of:

• cartels;
• abuse of dominance;
• exclusionary conduct;
• discriminatory conduct.

Layer 2 — Ex-ante digital regulation

Rules for particularly powerful platforms concerning:

• interoperability;
• access;
• self-preferencing;
• data portability.

Layer 3 — Algorithmic accountability

Requirements concerning:

• documentation;
• testing;
• monitoring;
• audit trails.

Layer 4 — Structural governance

Merger review and ecosystem-level intervention.

Layer 5 — Technical remedies

Including:

• APIs;
• interoperability;
• data portability;
• technical separation.

Layer 6 — International cooperation

Because AI commerce is inherently cross-border.

31. Role of Competition Authorities

Competition authorities may increasingly need specialist capabilities in:

• computer science;
• AI;
• data analytics;
• cybersecurity;
• machine learning;
• economics;
• platform architecture.

A future competition investigation could involve forensic analysis of an AI system rather than merely examination of emails and contracts.

32. Economic Analysis of Machine-Mediated Commerce

Traditional indicators such as market share and price remain relevant but may become insufficient.

Authorities may also consider:

Dynamic factors

• innovation;
• algorithmic learning;
• switching costs;
• network effects.

Technical factors

• API access;
• interoperability;
• data portability;
• computational infrastructure.

Behavioural factors

• automated steering;
• personalised pricing;
• ranking;
• recommendation.

Ecosystem factors

• integration across markets;
• platform dependence;
• cross-market data advantages.

33. The Problem of Explainability

A company may argue:

"The algorithm produced the outcome; no human employee instructed it to do so."

Competition law cannot simply accept this statement as a complete defence.

Otherwise, firms could potentially outsource important commercial decisions to autonomous systems and create a responsibility gap.

Future law may therefore require businesses to demonstrate:

• how the system was designed;
• what objectives were established;
• what safeguards existed;
• what monitoring was performed; and
• how harmful outcomes were addressed.

34. Compliance by Design

Competition compliance may increasingly need to be incorporated directly into software.

Examples include:

• prohibiting certain price-coordination inputs;
• preventing communication of competitively sensitive information;
• monitoring excessive price convergence;
• preventing discriminatory ranking;
• creating automatic alerts;
• preserving decision logs.

This can be described as:

Competition law by design.

Rather than investigating violations after they occur, systems could be designed to reduce the probability of violations from the beginning.

35. Challenges for Enforcement

35.1 Attribution

Who is responsible for an autonomous decision?

35.2 Causation

Did the algorithm actually cause the anticompetitive outcome?

35.3 Intent

Was coordination intended, foreseeable or accidental?

35.4 Explainability

Can the authority understand why the system produced the result?

35.5 Evidence

How should authorities obtain proprietary algorithmic evidence?

35.6 Jurisdiction

Which country regulates a transaction executed automatically across multiple jurisdictions?

35.7 Speed

Traditional investigations can take years, whereas algorithms can change market conditions within seconds.

36. Future Competition-Law Principles

Several principles are likely to become increasingly important.

1. Technological neutrality

Competition law should apply irrespective of whether conduct is performed manually or algorithmically.

2. Accountability

Businesses should remain responsible for systems they design and deploy.

3. Auditability

Important automated commercial decisions should leave sufficient records for investigation.

4. Interoperability

Dominant technological systems should not unnecessarily prevent competitive alternatives from functioning.

5. Non-discrimination

AI-mediated marketplaces should not arbitrarily discriminate against competing businesses.

6. Contestability

Markets should remain open to new competitors.

7. Data mobility

Control over data should not automatically become permanent control over the market.

37. Relationship Between Competition Law and AI Regulation

Competition law and AI regulation will increasingly overlap but should not be treated as identical.

AI regulation primarily asks:

Is the system safe, transparent, accountable and compliant?

Competition law asks:

Does the conduct restrict competition or strengthen market power unlawfully?

A single AI system can therefore raise both categories of questions.

38. Emerging Doctrine of Machine Accountability

A possible future doctrine could be built around the following proposition:

Autonomous operation should not create an autonomous exemption from competition law.

The relevant legal inquiry would examine:

1. who controlled the system;
2. who designed its objectives;
3. what information it received;
4. what decisions it was authorised to make;
5. whether harmful conduct was foreseeable;
6. whether monitoring existed;
7. whether corrective mechanisms existed; and
8. whether the undertaking benefited from the conduct.

This would avoid treating the machine itself as the legal decision-maker.

39. Future Remedies

Competition authorities may use increasingly sophisticated remedies.

Structural remedies

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

Behavioural remedies

• non-discrimination;
• prohibition of self-preferencing;
• restrictions on data use.

Technical remedies

• API access;
• interoperability;
• data portability;
• algorithmic auditing.

Procedural remedies

• reporting;
• compliance monitoring;
• independent audits;
• preservation of algorithmic records.

40. Overall Legal Framework

The future governance structure can therefore be expressed as:

Machine-Mediated Commerce

↓

Algorithmic Decision-Making

↓

Potential Competitive Effects

↓

Competition-Law Classification

→ Agreement / Concerted Practice
→ Abuse of Dominance
→ Vertical Restraint
→ Merger / Acquisition
→ Exclusionary Conduct
→ Discriminatory Conduct

↓

Algorithmic Evidence

↓

Competition Authority Investigation

↓

Technical + Behavioural + Structural Remedies

41. Key Case-Law Principles at a Glance

42. Conclusion

Machine-mediated commerce represents a transition from human-mediated markets to increasingly autonomous economic systems. The fundamental principles of competition law—prohibiting cartels, preventing abusive exclusion, protecting contestability and scrutinising concentrations—remain applicable, but their enforcement mechanisms must evolve.

The most significant future transformation will be the movement from reactive competition enforcement toward continuous technological governance.

Competition authorities will increasingly need to understand:

• algorithms;
• AI agents;
• data ecosystems;
• automated contracting;
• machine-to-machine transactions;
• platform architecture;
• interoperability; and
• algorithmic evidence.

The central legal principle should remain that automation changes the mechanism of commercial conduct, not the underlying obligation to comply with competition law. At the same time, future rules will need to distinguish genuine autonomous market adaptation from unlawful coordination and ensure that technological complexity does not create a gap between economic power and legal accountability.