Competition Law And Competition Governance In Autonomous Exchange Systems

1. Introduction

Autonomous Exchange Systems (AES) are technologically mediated systems in which exchanges between buyers, sellers, service providers, assets, or data can occur with limited or no direct human intervention. They may include:

• algorithmic trading and automated exchanges;
• digital asset and decentralized exchanges;
• autonomous procurement platforms;
• algorithmic marketplaces;
• smart-contract-based exchanges;
• AI-driven allocation and matching systems;
• automated pricing and bidding systems;
• machine-to-machine commercial transactions; and
• future systems in which AI agents independently negotiate and execute transactions.

“Autonomous exchange system” is not presently a single, universally recognized category of competition law. Instead, existing competition rules are applied to the economic conduct performed through the autonomous system. The central legal question is therefore not whether a machine made the decision, but whether the design, ownership, data architecture, algorithms, or human instructions produce conduct that restricts competition.

This distinction is particularly important because an autonomous system can make thousands of commercially significant decisions without a human approving each transaction. Competition law must therefore move from examining only individual human communications toward examining system architecture, data flows, incentives, governance mechanisms, and algorithmic decision-making.

2. Meaning Of Autonomous Exchange Systems

An autonomous exchange system can be represented as:

Participants → Data → Algorithm/AI → Matching/Allocation → Automated Transaction → Feedback → Further Decisions

For example:

Seller A → autonomous pricing agent → exchange protocol → Buyer B → smart contract → automatic settlement.

The system may determine:

• price;
• quantity;
• counterparty;
• timing;
• allocation;
• access;
• discounts;
• commissions;
• priority;
• routing;
• inventory;
• bidding;
• matching; and
• execution.

The more decision-making is delegated to software, the more difficult it becomes to identify traditional forms of collusion or exclusion.

3. Core Competition-Law Problem

Traditional competition law generally asks:

Did enterprises agree, coordinate, or independently behave in an anticompetitive manner?

Autonomous systems introduce additional questions:

1. Who designed the algorithm?
2. Who controls the algorithm?
3. Who supplies the training or market data?
4. Can competitors influence the algorithm?
5. Does the algorithm use competitors' confidential information?
6. Can participants independently deviate from its recommendations?
7. Does the system automatically punish deviations?
8. Does participation require accepting restrictive conditions?
9. Can competitors access the system on equal terms?
10. Can the system owner exclude competing systems?
11. Does the system create a bottleneck or essential facility?
12. Can the algorithm itself become a mechanism of coordination?

4. Legal Architecture

A. Agreement and Concerted Practice

Autonomous execution does not automatically eliminate the requirement for an agreement or concerted practice.

An enterprise cannot necessarily avoid cartel liability merely because:

“The algorithm, not the employee, fixed the price.”

Where humans previously agree to coordinate and software implements that agreement, ordinary cartel principles remain applicable.

This is particularly clear in United States v. Topkins, where algorithmic software was used to implement an online price-fixing arrangement.

5. Algorithmic Coordination

Autonomous systems can produce several different forms of coordination.

Model 1 — Human Agreement + Algorithmic Execution

Competitors agree:

“We will not price below X.”

Their software automatically enforces the agreement.

This is essentially a conventional cartel using technological implementation.

Model 2 — Platform-Mediated Coordination

Competitors use a common platform that provides pricing or allocation instructions.

The legal issue becomes whether the platform functions as a hub through which competitors coordinate.

Model 3 — Predictable Autonomous Reaction

Competitors do not expressly communicate, but each knows that the other's algorithm will react predictably to market movements.

This creates difficult questions concerning tacit coordination.

Model 4 — Common Data-Pooling Algorithm

Competitors provide confidential information to a common system.

The system then recommends prices or strategies.

This raises particularly serious competition concerns because the system can effectively transform competitors' private information into coordinated commercial decisions.

6. Six Major Case Laws

1. United States v. Topkins

Jurisdiction: United States
Subject: Online algorithmic price fixing

This is one of the most important cases for understanding algorithmically implemented cartels.

The defendants were involved in an agreement concerning prices for products sold online. Pricing software was configured to implement the agreed pricing strategy.

Principle

The use of software does not transform cartel conduct into lawful independent conduct.

The crucial distinction is between:

Independent algorithmic pricing

and

algorithmic implementation of an unlawful human agreement.

Relevance to Autonomous Exchange Systems

If autonomous trading agents are programmed pursuant to an agreement between competing enterprises, the autonomous character of execution does not provide immunity.

The competition authority can examine:

• source code;
• configuration instructions;
• communications;
• programming decisions;
• pricing rules;
• transaction records; and
• relationships between participants.

Thus:

Autonomous execution is not a defence to an underlying cartel.

2. Eturas UAB v. Lietuvos Respublikos konkurencijos taryba

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

This case concerned an online travel-booking platform through which travel agencies operated.

The platform administrator communicated a restriction concerning discounts, and the technological system subsequently limited the discounts available through the platform.

The CJEU examined whether participating businesses could be regarded as having engaged in concerted practice merely because the platform communicated the restriction.

Principle

The technological transmission of information does not automatically establish cartel participation. However, the surrounding circumstances and the conduct of participating undertakings after receiving the information can become relevant.

Importance

Eturas demonstrates that an online platform can become a mechanism through which competitively significant coordination takes place.

Application to Autonomous Exchanges

Suppose an autonomous exchange communicates:

“All participants must maintain a minimum transaction margin.”

The important questions would include:

• Did participants know about the restriction?
• Did they accept it?
• Did they continue participating?
• Did they have the ability to reject it?
• Did their subsequent behaviour demonstrate acceptance?
• Was the system designed to facilitate coordination?

Thus, autonomous systems require examination of both technology and participant conduct.

3. Trod Ltd and GB Eye Ltd v. Competition and Markets Authority

United Kingdom, Competition Appeal Tribunal, 2016

This case involved two online sellers using automated repricing software on Amazon Marketplace.

The businesses had agreed not to undercut one another, while software was used to implement the arrangement.

Principle

The fact that software automatically executes pricing decisions does not eliminate competition-law responsibility for the underlying agreement.

The software was essentially the mechanism for enforcing the agreement.

Significance

This case is particularly relevant to autonomous exchange systems because it illustrates the distinction between:

technology as a neutral commercial tool

and

technology deliberately configured to implement anti-competitive coordination.

Autonomous-System Application

An autonomous exchange system that automatically prevents a participant from undercutting another participant could attract scrutiny if the restriction results from an anti-competitive agreement or coordinated strategy.

7. 4. Samir Agrawal v. Competition Commission of India

Supreme Court of India, 2021

This is particularly important for Indian competition-law analysis of algorithmic marketplaces.

The dispute concerned allegations that Ola and Uber used algorithmic pricing mechanisms that effectively prevented drivers from independently determining fares.

The allegation was that the platforms could function as a hub, with drivers acting as spokes.

The CCI did not find a prima facie cartel, and the Supreme Court ultimately upheld the relevant outcome. The courts emphasized the absence of sufficient evidence of an agreement or common hub-based exchange of competitively sensitive information between the relevant participants.

Importance

The case establishes an important conceptual limitation:

Algorithmic price similarity or algorithmic pricing, by itself, does not necessarily establish collusion.

Independent algorithms can respond similarly to:

• demand;
• supply;
• traffic;
• weather;
• events;
• time;
• consumer behaviour; and
• other market variables.

The CCI specifically distinguished algorithmic pricing from a conventional hub-and-spoke arrangement where a third party facilitates the exchange of competitively sensitive information.

Relevance to Autonomous Exchange Systems

An autonomous exchange should therefore not be presumed unlawful simply because:

multiple autonomous agents produce similar prices.

The authority must investigate why the similarity exists.

Possible explanations include:

• independent optimization;
• common market conditions;
• common public data;
• common software;
• confidential information sharing;
• common instructions;
• coordinated programming; or
• deliberate collusion.

This creates a major evidentiary distinction in autonomous markets.

8. 5. RealPage Antitrust Litigation

United States — algorithmic rental-pricing litigation

RealPage concerned allegations surrounding algorithmic pricing software used by landlords.

The central competition concern was that landlords could provide competitively sensitive information to a common pricing system, which would then generate recommendations affecting rental prices.

The U.S. Department of Justice's position highlighted the possibility that an algorithm using competitors' non-public information could facilitate unlawful coordination even where competitors did not communicate directly with one another. The matter subsequently resulted in a settlement involving restrictions on the use of competitors' confidential information in rental pricing recommendations.

Significance

RealPage is important because it moves beyond the simple:

“human agreement → algorithm executes”

model.

Instead, the alleged mechanism can be conceptualized as:

Competitor A data + Competitor B data + Common algorithm → coordinated pricing environment

Autonomous Exchange Application

This is particularly significant for AI-driven exchanges.

If competing autonomous agents continually upload:

• prices;
• inventories;
• costs;
• demand forecasts;
• capacity;
• margins; and
• strategic information

to a common AI system, the system could potentially reduce strategic uncertainty between competitors.

The competition concern therefore lies not merely in the final price but in the information architecture underlying the price.

9. 6. Meru Travel Solutions Pvt. Ltd. v. Uber India Systems Pvt. Ltd.

Competition Commission of India, 2021

Meru challenged Uber's conduct in the radio-taxi market, including allegations involving dominant-position conduct and pricing strategies.

The case is relevant because it demonstrates that competition analysis of technology platforms cannot be restricted to cartel questions. A technologically autonomous platform may also be examined under abuse-of-dominance principles.

The CCI considered allegations concerning Uber's position in the relevant market and alleged predatory pricing.

Importance for Autonomous Exchanges

An autonomous exchange system can become problematic even without a cartel.

Suppose a dominant autonomous exchange controls access to a market and its algorithm:

• excludes competing exchanges;
• discriminates against rival participants;
• gives preferential access to affiliated entities;
• imposes unreasonable access conditions;
• uses transaction data to disadvantage rivals; or
• engages in exclusionary pricing.

The legal issue may then shift from Section 3-type coordination to abuse of dominance.

10. Additional Important Authorities

Although the six cases above are particularly useful, several other authorities strengthen the framework.

Competition Commission of India v. Steel Authority of India Ltd.

The Supreme Court's approach to market power, relevant markets and abuse of dominance provides the broader Indian doctrinal framework within which technologically mediated exchange systems can be assessed.

Excel Crop Care Ltd. v. CCI

This case is important for cartel analysis and penalty principles in India. It illustrates the broader proposition that competition authorities examine the economic substance of coordinated conduct rather than merely its formal structure.

Rajasthan Cylinders and Containers Ltd. v. Union of India

The case demonstrates the importance of evidence when determining whether parallel commercial conduct actually constitutes cartelization.

These authorities become relevant when an autonomous exchange produces apparently coordinated outcomes but the authority must determine whether there is sufficient evidence of actual coordination.

11. Autonomous Exchange Systems And Abuse Of Dominance

An autonomous exchange may become a digital bottleneck.

Consider a platform controlling:

users + data + liquidity + matching + settlement + identity + transaction history

Such integration can create significant market power.

Potential exclusionary practices include:

1. Self-preferencing

The exchange gives its own trading agent priority over independent agents.

2. Discriminatory access

Competitors receive slower or inferior access to APIs.

3. Data foreclosure

The exchange prevents competing systems from obtaining necessary transaction data.

4. Algorithmic exclusion

The matching algorithm systematically deprioritizes rival suppliers.

5. Loyalty mechanisms

The system rewards participants who exclusively use the platform.

6. Interoperability restrictions

The exchange refuses to communicate with competing exchanges.

7. Predatory algorithmic pricing

The dominant operator's system deliberately prices below an appropriate cost benchmark to eliminate rivals.

12. Essential-Facility Issues

Some autonomous exchanges could evolve into infrastructure through which competitors must transact.

For example:

Autonomous electricity exchange → automated matching → settlement → grid access.

Or:

Digital-asset exchange → liquidity → settlement → identity → transaction data.

If a system becomes sufficiently indispensable, competition law may raise questions concerning:

• access;
• interoperability;
• non-discrimination;
• refusal to deal;
• data portability;
• API access;
• technical standards; and
• interoperability with rival systems.

However, being technologically important does not automatically make a system an essential facility. Traditional legal requirements concerning indispensability, competition effects and objective justification remain relevant.

13. Autonomous Exchanges And Data

Data may be the most important competition resource in an autonomous exchange.

An exchange can accumulate:

• transaction histories;
• real-time prices;
• consumer preferences;
• inventory;
• liquidity;
• bids and offers;
• trading strategies;
• identity information;
• credit information; and
• competitor behaviour.

This produces a potential data feedback loop:

More users
↓
More transactions
↓
More data
↓
Better algorithm
↓
Better matching/pricing
↓
More users

The resulting network effect can create substantial barriers to entry.

14. Data Pooling Between Competitors

Competition law should distinguish:

Legitimate data sharing

For example:

• anonymized market statistics;
• regulatory reporting;
• technical interoperability;
• fraud prevention.

from:

Potentially problematic data sharing

For example:

• future prices;
• individual competitor margins;
• planned capacity;
• customer-specific information;
• strategic inventory information.

An autonomous system that aggregates competitively sensitive information from competitors can reduce strategic uncertainty and potentially facilitate coordination.

15. Autonomous Agents As Competitors

A particularly difficult future question is:

Can autonomous AI agents themselves become competitors?

Suppose:

• AI Agent A represents Company A;
• AI Agent B represents Company B;
• both negotiate continuously;
• each agent independently determines price;
• transactions are automatically executed.

Competition law should ordinarily look through the technological mechanism to the economic actors controlling or benefiting from the agents.

The existence of an autonomous agent does not automatically create a new legal person or remove the responsibility of its operator.

16. Machine-to-Machine Commerce

Autonomous exchange systems could eventually operate without direct consumer involvement.

For example:

AI manufacturer → autonomous procurement agent → autonomous supplier agent → automated contract → automated payment

The transaction could occur within milliseconds.

This raises new competition questions:

• Can machines collude?
• Can machines tacitly coordinate?
• Who is responsible for their conduct?
• What if the algorithm independently discovers supra-competitive prices?
• What if the algorithm learns to punish deviations?
• Can the system administrator be responsible for foreseeable algorithmic behaviour?

The legal focus will likely remain on the human or corporate governance structure behind the machine rather than treating the machine as an independent economic actor.

17. Algorithmic Tacit Coordination

One of the most difficult issues is tacit algorithmic coordination.

Imagine:

Firm A's algorithm raises price → Firm B's algorithm observes it → B raises price → A observes B → A maintains price.

No explicit agreement may exist.

The difficulty is distinguishing:

legitimate intelligent adaptation

from

deliberate algorithmic coordination.

Evidence might include:

• source code;
• training instructions;
• system objectives;
• communications with software developers;
• historical data;
• algorithmic constraints;
• monitoring functions;
• punishment mechanisms;
• unusual price responses; and
• evidence that the system was intentionally designed to coordinate.

18. Competition Governance

Competition governance for autonomous exchanges should operate at several levels.

Level 1 — Human Governance

Boards and senior management should determine:

• permissible algorithmic objectives;
• competition-law risk tolerances;
• data-sharing policies;
• access rules;
• audit requirements; and
• escalation procedures.

Level 2 — Algorithmic Governance

Algorithms should be designed with:

• competition-law constraints;
• auditability;
• explainability;
• independent decision-making;
• access neutrality;
• data minimization; and
• safeguards against coordinated behaviour.

Level 3 — Data Governance

There should be clear controls over:

• collection;
• storage;
• sharing;
• aggregation;
• anonymization;
• retention;
• competitor information; and
• cross-platform data flows.

Level 4 — Market Governance

The exchange should establish rules concerning:

• participation;
• access;
• pricing;
• matching;
• interoperability;
• dispute resolution;
• suspension;
• technical standards; and
• algorithmic transparency.

Level 5 — Regulatory Governance

Competition authorities may need:

• algorithmic audit powers;
• data-access powers;
• technical investigative expertise;
• computational evidence procedures;
• real-time monitoring;
• sandbox mechanisms; and
• coordination with financial, telecom, data-protection and sector regulators.

19. Competition Risks — A Taxonomy

20. Evidentiary Problems

Autonomous systems create a major change in competition-law evidence.

Traditional evidence includes:

• emails;
• meetings;
• contracts;
• telephone records; and
• witness testimony.

Autonomous systems additionally generate:

• source code;
• model weights;
• prompts;
• system instructions;
• API logs;
• transaction logs;
• training datasets;
• telemetry;
• model outputs;
• audit trails;
• version histories; and
• automated decision records.

Consequently, competition investigations may increasingly become technical forensic investigations.

21. The "Black Box" Problem

A company might argue:

“We cannot explain why the AI produced this result.”

That explanation becomes increasingly difficult to accept where the enterprise controls:

• the system architecture;
• training data;
• objectives;
• deployment environment;
• monitoring mechanisms; and
• commercial incentives.

The legal importance of explainability therefore increases as autonomy increases.

A useful governance principle is:

The greater the commercial autonomy of the system, the greater the need for traceability and auditability.

22. Regulatory Sandbox Model

A useful competition-governance framework could be:

Stage 1 — Registration

Identify autonomous systems capable of materially affecting competition.

Stage 2 — Risk Classification

Classify systems according to:

• market share;
• degree of autonomy;
• data sensitivity;
• network effects;
• switching costs;
• market criticality.

Stage 3 — Pre-Deployment Review

Review:

• pricing mechanisms;
• access rules;
• data architecture;
• interoperability;
• exclusivity provisions.

Stage 4 — Continuous Monitoring

Monitor:

• prices;
• access;
• concentration;
• discriminatory outcomes;
• algorithmic convergence.

Stage 5 — Independent Audit

Conduct periodic technical and competition audits.

Stage 6 — Corrective Measures

Possible remedies include:

• data separation;
• interoperability;
• access obligations;
• algorithm modification;
• non-discrimination rules;
• divestiture; or
• behavioural commitments.

23. Key Legal Principles Emerging From The Case Law

The six principal cases collectively support several propositions.

Principle 1

Technology does not immunize anti-competitive conduct.

Topkins and Trod demonstrate this particularly clearly.

Principle 2

Algorithmic pricing alone is not necessarily collusion.

Samir Agrawal illustrates the importance of evidence of an agreement or coordinated mechanism.

Principle 3

A common technological intermediary can facilitate coordination.

Eturas demonstrates the importance of platform-mediated communication and participant conduct.

Principle 4

Common algorithms using competitors' sensitive information present heightened risk.

RealPage illustrates this problem.

Principle 5

Autonomous systems may raise unilateral-conduct issues as well as cartel issues.

The Uber/Meru litigation demonstrates the relevance of dominance analysis.

Principle 6

Competition law must examine economic substance rather than merely the technical form of the transaction.

An autonomous exchange is therefore not outside competition law merely because the transaction is executed by code.

24. Future Competition-Law Framework

A future legal framework for autonomous exchanges can be expressed as:

AUTONOMOUS EXCHANGE

↓

Who controls the system?

↓

What market does it affect?

↓

What data does it receive?

↓

Does it receive competitors' sensitive information?

↓

Can participants independently deviate?

↓

Does the system facilitate coordination?

↓

Does the operator possess substantial market power?

↓

Does the system exclude competitors?

↓

Does it discriminate between participants?

↓

Are interoperability and access available?

↓

Are the system's decisions auditable?

↓

Competition-law assessment

25. Autonomous Decentralized Exchanges

Decentralized systems introduce an additional problem.

There may be:

• no conventional intermediary;
• no central pricing authority;
• distributed governance;
• smart contracts;
• token holders;
• automated market makers; and
• geographically dispersed participants.

The absence of a traditional corporate intermediary does not necessarily mean absence of competition-law issues.

Authorities may instead need to identify:

• developers;
• governance organizations;
• protocol operators;
• token holders;
• liquidity providers;
• front-end operators; and
• entities exercising effective control.

The central question becomes:

Where does economically meaningful control reside?

26. Competition Governance And Smart Contracts

Smart contracts can automatically enforce:

• exclusivity;
• minimum prices;
• transaction fees;
• access conditions;
• allocation rules;
• loyalty incentives; and
• settlement restrictions.

This produces a significant tension:

Smart contract = certainty of execution

but potentially:

Smart contract = rigidity of anti-competitive restriction.

Competition compliance should therefore occur before deployment, because changing an immutable or widely distributed protocol may be technically difficult after adoption.

27. Remedies

Competition authorities may increasingly require technologically specific remedies.

Behavioural remedies

• prohibit sensitive-data inputs;
• modify algorithmic objectives;
• prohibit discriminatory access;
• require independent audits.

Structural remedies

• separate exchange and trading operations;
• separate data infrastructure;
• separate affiliated autonomous agents.

Interoperability remedies

• API access;
• common technical standards;
• data portability;
• protocol interoperability.

Transparency remedies

• algorithmic documentation;
• audit trails;
• decision logs;
• governance disclosures.

Monitoring remedies

• continuous compliance monitoring;
• independent technical auditors;
• periodic competition assessments.

28. Compliance Framework For Enterprises

An enterprise operating an autonomous exchange should implement:

1. Algorithm inventory
2. Competition-risk classification
3. Sensitive-data controls
4. Competitor-information firewall
5. Algorithmic independence testing
6. Human oversight
7. Audit logs
8. Version control
9. Red-team testing
10. Periodic competition-law audits
11. Incident reporting
12. Board-level accountability

A particularly important rule should be:

No autonomous system should be permitted to use competitors' competitively sensitive information unless the legal basis and competitive effects have been carefully assessed.

29. Exam-Oriented Critical Analysis

The principal challenge is that traditional competition law was developed around human decision-making, whereas autonomous exchange systems increasingly involve machine-mediated decision-making.

Nevertheless, the basic economic objectives remain recognizable:

• preventing collusion;
• preserving independent decision-making;
• preventing exclusion;
• protecting market access;
• preserving innovation;
• preventing exploitation of market power; and
• maintaining contestability.

The principal doctrinal challenge is therefore not necessarily the creation of an entirely new competition law, but the development of new evidentiary, technical and governance mechanisms for applying existing competition principles to autonomous commercial systems.

30. Conclusion

Competition law in autonomous exchange systems is likely to develop around a fundamental proposition:

Autonomy changes the mechanism of market behaviour, but it does not eliminate competition-law responsibility.

The most important distinctions are between:

independent algorithmic behaviour
and
algorithmically implemented coordination;

efficient automated matching
and
algorithmic exclusion;

legitimate data aggregation
and
competitively sensitive data pooling;

and

decentralized technological architecture
and
effective economic control.

The jurisprudence of Topkins, Eturas, Trod, Samir Agrawal, RealPage and Meru provides a useful foundation. Together, these authorities show why future competition governance must combine traditional antitrust principles with algorithmic auditing, data governance, interoperability, technical evidence and continuous regulatory supervision.

Ultimately, the central legal question for an autonomous exchange system will be: