Competition Law and Future Competition Frameworks for Autonomous Commerce

1. Introduction

Autonomous commerce refers to markets in which software agents, AI systems, autonomous purchasing assistants, smart contracts, or machine-to-machine systems can independently search for products, compare prices, negotiate terms, select suppliers, place orders, make payments, manage inventory, and sometimes renegotiate contracts.

The development of agentic AI is particularly important because competition law has traditionally assumed that human firms make commercial decisions. In autonomous commerce, an AI agent may make thousands of commercially significant decisions without a human approving each transaction.

The OECD has specifically identified agentic AI, algorithmic collusion, data access, model restrictions, vertical integration and attribution of liability as emerging competition-law issues. It also notes that AI can simultaneously lower entry barriers and create new forms of market power.

The central future question is therefore:

How should competition law regulate markets where commercial decisions are increasingly made by autonomous computational agents rather than directly by humans?

2. Meaning and Characteristics of Autonomous Commerce

Autonomous commerce can involve:

• AI shopping agents;
• autonomous procurement systems;
• algorithmic price negotiation;
• machine-to-machine contracting;
• autonomous inventory management;
• AI-powered marketplaces;
• automated advertising purchases;
• autonomous financial transactions;
• smart-contract commerce;
• AI-generated product recommendations;
• autonomous supply-chain decisions;
• dynamic and personalised pricing;
• automated switching between suppliers.

For example, a company could instruct an AI procurement agent:

"Maintain 10,000 units of inventory at the lowest economically viable total cost."

The agent could then independently:

1. identify suppliers;
2. compare prices;
3. evaluate quality;
4. negotiate;
5. switch suppliers;
6. sign a digital contract;
7. make payment;
8. reorder automatically.

This creates a competition-law problem because the commercial actor is no longer simply a human manager—it is an autonomous decision system operating within an ecosystem of other decision systems.

3. Why Traditional Competition Law Faces Difficulty

Traditional competition law generally focuses on:

A. Agreements

Competitors must not agree to:

• fix prices;
• allocate markets;
• restrict output;
• exchange competitively sensitive information.

B. Abuse of dominance

Dominant undertakings must not:

• exclude competitors;
• discriminate without justification;
• impose unfair conditions;
• engage in tying or bundling;
• deny access to essential inputs;
• exploit network effects unlawfully.

C. Merger control

Authorities examine whether a transaction substantially reduces competition.

D. Consumer welfare and competitive process

Authorities traditionally examine:

• prices;
• output;
• quality;
• innovation;
• choice;
• entry conditions.

Autonomous commerce complicates each category.

4. Major Competition Problems in Autonomous Commerce

I. Algorithmic Collusion

The most immediate concern is that autonomous agents may independently learn that maintaining higher prices is commercially advantageous.

Two competing agents might:

• observe one another's prices;
• repeatedly interact;
• predict each other's behaviour;
• automatically respond to deviations;
• converge on stable prices.

There may be no telephone call, email or traditional cartel meeting.

This creates the distinction between:

Explicit collusion

Humans instruct the systems to collude.

Facilitated collusion

A common platform or algorithm facilitates coordination.

Autonomous algorithmic coordination

Machines independently learn strategies that reduce competitive rivalry.

The first two categories fit relatively comfortably within existing law. The third presents the more difficult future question of whether competition law should impose responsibility where there is no conventional human agreement.

The OECD has identified common pricing software and common model providers as particular risks for algorithmic coordination.

5. Common AI Providers as "Digital Hubs"

Suppose ten competing retailers use the same AI provider.

The AI provider receives:

• prices;
• inventory;
• sales data;
• customer information;
• demand forecasts;
• discounts;
• promotional plans.

It then generates recommendations for each retailer.

The AI provider can potentially become a digital hub connecting competing firms.

This creates a modern version of the hub-and-spoke problem.

The competition-law question becomes:

When does use of a common autonomous system become unlawful coordination rather than ordinary use of technology?

This issue is particularly significant because the algorithm may itself determine what information is collected and how recommendations are generated.

6. Autonomous Price Discrimination

AI agents can calculate different prices for different consumers based on:

• purchasing history;
• location;
• willingness to pay;
• browsing behaviour;
• urgency;
• inventory conditions;
• income proxies;
• device characteristics;
• previous negotiations.

Autonomous commerce could therefore create real-time personalised pricing.

Competition law must distinguish between:

Legitimate dynamic pricing

Prices change according to genuine supply and demand.

Competitive price discrimination

Different prices reflect legitimate commercial segmentation.

Anticompetitive discrimination

Algorithmic discrimination may exclude rivals, exploit market power or disadvantage particular trading partners.

7. Autonomous Self-Preferencing

An AI marketplace may control:

1. search;
2. ranking;
3. recommendations;
4. advertising;
5. payments;
6. fulfilment;
7. seller access.

If the platform also sells its own products, its autonomous ranking system could automatically favour those products.

This creates a future form of machine-generated self-preferencing.

The problem is particularly significant because the platform could claim:

"The algorithm independently determined the ranking."

Competition law cannot necessarily treat algorithmic autonomy as a complete defence.

The relevant question is likely to remain:

Who designed the system, what objectives were embedded into it, what information did it use, and what competitive effects resulted?

8. Autonomous Agents and Data Advantages

Autonomous commerce depends heavily upon data.

An established platform may possess:

• consumer transaction data;
• supplier data;
• competitor prices;
• demand forecasts;
• logistics information;
• advertising data;
• product performance data.

An AI system can convert those datasets into increasingly powerful commercial predictions.

This creates a potential data-feedback loop:

More transactions → more data → better AI → better predictions → more transactions → still more data

Such loops can make entry increasingly difficult.

9. Interoperability and Switching

Autonomous commerce requires agents to communicate.

Different agents may use different:

• APIs;
• protocols;
• identity systems;
• payment systems;
• data formats;
• AI models.

A dominant platform could potentially restrict interoperability.

Examples include:

• refusing API access;
• charging discriminatory access fees;
• limiting agent compatibility;
• preventing competitors' AI agents from accessing marketplace information;
• blocking automated switching.

Future competition law therefore needs to treat agent interoperability as potentially analogous to access to other technologically important infrastructure.

10. Autonomous Procurement and Buyer Power

Autonomous commerce does not only create seller-side dominance.

Large purchasers may deploy AI procurement systems capable of simultaneously negotiating with thousands of suppliers.

This may create substantial algorithmic buyer power.

A large buyer's AI might:

• automatically demand discounts;
• compare suppliers in real time;
• threaten automatic switching;
• coordinate procurement across subsidiaries;
• impose standardised contract terms.

The competition question becomes whether autonomous procurement increases efficiency or creates abusive monopsony/oligopsony power.

11. Six Important Case Laws

1. United States v. Topkins

United States v. Topkins, No. CR 15-00201 (N.D. Cal. 2015) is one of the clearest early examples of algorithm-assisted price fixing.

Online sellers agreed to coordinate prices for posters sold through Amazon Marketplace and used automated pricing software to implement the arrangement.

Significance

The case demonstrates that:

The use of an algorithm does not transform an unlawful human agreement into lawful competition.

The software was effectively the mechanism through which the agreed pricing strategy was implemented.

Future relevance

Autonomous commerce will require authorities to determine whether an AI agent is:

• merely implementing an agreement;
• facilitating coordination;
• or independently generating anticompetitive behaviour.

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

The UK case involving Trod Ltd. and GB Eye Ltd. concerned online sellers using automated repricing software.

The parties had agreed to maintain minimum prices, with software helping implement the arrangement.

Significance

The case establishes an important principle for autonomous commerce:

automation does not eliminate responsibility for an underlying anticompetitive arrangement.

Future application

Where autonomous purchasing or pricing agents execute an unlawful commercial strategy, competition authorities may examine:

• who programmed the agent;
• what instructions were provided;
• whether competitors coordinated;
• whether sensitive information was exchanged;
• whether the resulting conduct was foreseeable.

3. Eturas UAB v. Lietuvos Respublikos konkurencijos taryba

Case C-74/14, Eturas UAB v. Lietuvos Respublikos konkurencijos taryba concerned an online travel-booking platform and a system message restricting discounts available through the platform.

The CJEU considered whether participants could be treated as having participated in coordinated conduct through the platform.

Significance

The case is extremely important for autonomous commerce because it demonstrates that:

a digital platform can become the mechanism through which competitors coordinate behaviour.

But participation cannot simply be presumed merely because a company uses a platform.

Future lesson

Autonomous commerce requires evidence concerning:

• knowledge;
• participation;
• acceptance;
• communication;
• implementation;
• distancing from the conduct.

4. Samir Agarwal v. Competition Commission of India

Samir Agarwal v. Competition Commission of India, (2021) 3 SCC 136 concerned allegations relating to pricing and the role of an intermediary/platform in facilitating coordination.

The case is important to the developing Indian understanding of hub-and-spoke arrangements.

Significance for autonomous commerce

An AI platform could become a technologically sophisticated hub connecting competing commercial actors.

The competition analysis may therefore ask:

• Did the hub transmit competitively sensitive information?
• Did competitors knowingly participate?
• Was there an agreement or concerted arrangement?
• Did the intermediary facilitate coordinated pricing?

This framework becomes increasingly important when AI systems replace traditional human intermediaries.

5. United States v. RealPage, Inc.

United States and State Plaintiffs v. RealPage, Inc. represents one of the most significant modern algorithmic-pricing disputes.

The U.S. Department of Justice alleged that landlords supplied non-public competitively sensitive information to RealPage's pricing system and that the software generated rental pricing recommendations based upon that information.

The subsequent settlement framework addressed the use of competitors' sensitive information and algorithmic pricing practices.

Significance

RealPage demonstrates that competition authorities can focus on:

• data inputs;
• algorithmic recommendations;
• common software;
• information sharing;
• alignment of competitors' conduct.

Autonomous-commerce lesson

The critical legal question is unlikely to be simply:

"Was AI used?"

Instead, it is likely to be:

What competitive information did the AI use, who supplied it, how was it processed, and how did the resulting system affect independent competitive decision-making?

6. Google Shopping

The European Commission's Google Shopping decision, followed by General Court litigation, is highly relevant to autonomous commerce because it concerns algorithmically organised search results and the preferential treatment of a platform's own comparison-shopping service.

Significance

The case illustrates the competition implications of:

• ranking algorithms;
• platform gatekeeping;
• self-preferencing;
• control over visibility;
• leveraging dominance from one market into another.

Future application

In autonomous commerce, the equivalent problem could arise where an AI shopping agent decides which:

• products to display;
• suppliers to contact;
• payment system to use;
• logistics provider to select;
• marketplace to access.

The more commercially important the agent becomes, the greater the importance of transparent and competitively neutral ranking criteria.

12. Additional Relevant Precedent: RealPage-Type Algorithmic Coordination

Recent enforcement developments show that algorithmic coordination is no longer purely theoretical.

The U.S. Department of Justice's RealPage proceedings have included settlements and proposed judgments involving landlords accused of using algorithmic pricing systems and competitively sensitive information.

This demonstrates an important future direction:

competition remedies may increasingly regulate the architecture and data inputs of algorithms rather than merely prohibiting the final price.

13. Future Competition Framework I — Algorithm Accountability

Competition authorities should develop an Algorithmic Competition Accountability Framework.

Every commercially significant autonomous agent could be required to maintain:

• model documentation;
• decision logs;
• training-data provenance;
• pricing rules;
• objective functions;
• material changes;
• external data sources;
• API connections;
• human override mechanisms.

This would allow regulators to reconstruct:

Input → Model → Decision → Market Effect

14. Future Framework II — Agentic Antitrust Liability

A future legal framework could distinguish three levels:

Level 1 — Human-directed conduct

A person deliberately instructs the AI to engage in anticompetitive conduct.

Level 2 — AI-facilitated conduct

Humans establish the arrangement and AI implements it.

Level 3 — Autonomous conduct

The AI independently develops a strategy that produces anticompetitive effects.

The first two categories fit relatively comfortably within existing doctrine.

The third presents the greatest legal challenge.

15. Future Framework III — Safe Harbour for Independent Algorithms

Competition law should avoid treating every common algorithm as inherently unlawful.

A safe-harbour framework could potentially protect autonomous systems where:

• competitors do not exchange confidential information;
• models are independently trained;
• inputs are not competitor-specific;
• pricing decisions remain independent;
• the provider does not coordinate competitors;
• the system does not intentionally suppress competitive responses.

This would preserve technological innovation while addressing genuine coordination risks.

16. Future Framework IV — AI Interoperability Rules

Competition authorities may increasingly need powers to require dominant platforms to provide:

• API access;
• data portability;
• agent interoperability;
• technical documentation;
• reasonable authentication mechanisms;
• non-discriminatory access.

The objective would be to prevent dominant ecosystems from becoming closed autonomous-commerce environments.

17. Future Framework V — Competition-by-Design

Instead of investigating anticompetitive behaviour only after harm occurs, competition law could move toward competition-by-design.

AI developers could be expected to incorporate:

• anti-collusion safeguards;
• information firewalls;
• independent pricing modules;
• audit trails;
• competitive neutrality;
• data minimisation;
• switching functionality.

This would be analogous to privacy-by-design concepts but focused on competitive markets.

18. Future Framework VI — Autonomous Merger Control

Autonomous commerce could also transform merger analysis.

Traditional merger review examines:

• market shares;
• concentration;
• entry barriers;
• efficiencies;
• innovation.

Future analysis may additionally examine:

AI capability

Can the combined company create a substantially superior autonomous commercial agent?

Data concentration

Will the transaction combine uniquely valuable datasets?

Ecosystem integration

Will the transaction combine:

AI model + marketplace + payments + logistics + advertising + consumer data?

Agent dependency

Will independent businesses become dependent on one company's commercial AI infrastructure?

This may require competition authorities to examine ecosystem concentration, rather than simply traditional product-market shares.

19. Future Framework VII — Competition Audits

Large autonomous-commerce systems could periodically undergo independent competition audits.

A competition audit could examine:

20. Autonomous Commerce and Essential Facilities

A powerful autonomous-commerce platform could become an important commercial gateway.

For example:

AI agent → dominant marketplace → payment network → logistics system

If competitors cannot reasonably reach consumers without access to that ecosystem, traditional doctrines concerning:

• essential facilities;
• refusal to deal;
• interoperability;
• access obligations

may become increasingly important.

The challenge will be determining when an AI ecosystem is sufficiently indispensable to justify intervention.

21. Autonomous Commerce and Consumer Choice

Competition law should also preserve consumer agency.

An autonomous purchasing agent could make decisions that consumers never directly see.

For example:

Consumer asks for "a suitable laptop under ₹80,000."

The agent may silently determine:

• which marketplaces are searched;
• which brands are considered;
• what commissions are relevant;
• whether sponsored products are included;
• what personal data is used;
• what products are excluded.

This creates a future distinction between:

consumer choice and machine-mediated choice.

Competition law may therefore need transparency regarding commercially significant recommendation criteria.

22. The Problem of Explainability

Traditional enforcement often asks:

"Why did the undertaking behave this way?"

With autonomous AI, the answer may be:

"The model generated that decision."

That cannot automatically end the inquiry.

Competition authorities may need access to:

• decision logs;
• model documentation;
• training data;
• reward functions;
• prompts;
• system instructions;
• model versions;
• transaction histories.

The OECD has specifically identified attribution of liability and agentic AI as areas requiring further competition-policy research.

23. Jurisdictional Problems

Autonomous commerce is inherently cross-border.

An AI agent could be:

• developed in the United States;
• hosted in Singapore;
• operated by an Indian company;
• purchasing from a German supplier;
• selling to consumers in Denmark;
• using a payment system in Ireland.

Consequently, competition authorities will face questions concerning:

• territorial jurisdiction;
• extraterritorial effects;
• evidence gathering;
• cross-border algorithm audits;
• conflicting regulatory requirements;
• international cooperation.

This suggests a need for greater convergence among:

• EU competition law;
• U.S. antitrust law;
• Indian competition law;
• UK competition law;
• OECD principles;
• other digital-market regimes.

24. Proposed Future Model

A comprehensive future framework could be represented as:

Autonomous Commerce

↓

1. Market Mapping

Identify agents, platforms, data providers and infrastructure.

↓

2. Algorithm Classification

Determine whether the system is:

• independent;
• intermediary;
• coordinating;
• dominant;
• vertically integrated.

↓

3. Data Examination

Identify:

• proprietary data;
• competitor data;
• consumer data;
• commercially sensitive information.

↓

4. Conduct Assessment

Examine:

• pricing;
• ranking;
• access;
• tying;
• self-preferencing;
• discrimination;
• exclusion;
• information exchange.

↓

5. Competitive Effects

Assess:

• prices;
• output;
• innovation;
• entry;
• choice;
• quality;
• switching.

↓

6. Accountability

Identify responsibility among:

• developer;
• platform;
• seller;
• user;
• AI provider;
• data provider.

↓

7. Remedy

Possible remedies include:

• behavioural commitments;
• interoperability;
• data separation;
• algorithm modification;
• access obligations;
• monitoring;
• structural remedies.

25. Key Future Legal Principles

A mature autonomous-commerce competition regime is likely to require the following principles:

1. Algorithmic neutrality

AI should not automatically receive immunity merely because conduct is machine-generated.

2. Human responsibility

Businesses should remain accountable for commercially significant systems they deploy.

3. Technological neutrality

Competition law should apply regardless of whether conduct occurs through humans, software, AI agents or smart contracts.

4. Data neutrality

Competitively sensitive information should not become a mechanism for automated coordination.

5. Interoperability

Dominant autonomous ecosystems should not unnecessarily prevent rival agents from competing.

6. Auditability

Authorities should be able to reconstruct important autonomous decisions.

7. Contestability

Markets should remain open to new AI agents and competitors.

8. Proportionality

Innovation-enhancing AI should not be prohibited merely because it creates efficiency or sophisticated pricing.

26. Major Doctrinal Challenges

The future debate will likely centre on several unresolved questions:

1. Can an AI agent itself constitute an "agreement"?
2. Who is legally responsible for autonomous collusion?
3. Can tacit algorithmic coordination violate competition law without human communication?
4. When does common AI infrastructure become a hub-and-spoke arrangement?
5. Should dominant AI agents have interoperability obligations?
6. Can autonomous agents engage in unlawful self-preferencing?
7. How should competition authorities audit black-box models?
8. Can algorithmic market power exist without traditional high market share?
9. Should AI providers be treated as infrastructure providers in some markets?
10. How should merger control address combinations of AI, data and marketplaces?

27. Conclusion

Autonomous commerce represents a fundamental transformation of competition rather than merely another application of e-commerce.

The historical cases involving Topkins, Trod/GB Eye and Eturas establish that technology does not automatically remove competition-law responsibility. Samir Agarwal demonstrates the continuing importance of intermediary and hub-and-spoke analysis, while RealPage illustrates the modern regulatory focus on algorithmic pricing, common software and competitively sensitive information. Google Shopping demonstrates how algorithmic ranking and platform control can become competition-law issues.

The future framework should therefore move from a narrow question of:

"Did humans expressly agree to restrict competition?"

toward a broader technological inquiry:

"How was the autonomous system designed, what information and objectives shaped its decisions, how did it affect independent competitive behaviour, and who controlled the relevant commercial infrastructure?"

The appropriate future model is likely to combine traditional antitrust rules + algorithmic auditing + data governance + interoperability + competition-by-design + AI-specific accountability.

Autonomous commerce should not be treated as inherently anticompetitive. AI agents can also reduce search costs, lower barriers to entry, improve procurement, increase price transparency and facilitate product differentiation. The OECD accordingly emphasises that the competitive effects of agentic AI are context-dependent and can be both pro-competitive and harmful to contestability.

The central objective of future competition law will therefore be to ensure that autonomous decision-making increases market efficiency without allowing autonomous systems to become mechanisms for cartelisation, exclusion, discrimination or the consolidation of unchallengeable digital market power.