COMPETITION LAW AND COMPETITION LAW ADAPTATION TO AUTONOMOUS ENTERPRISES

1. Introduction

Competition law traditionally assumes that enterprises are controlled by human managers who consciously make decisions about price, output, supply, distribution, investment, contracts and market strategy.

The development of artificial intelligence, machine learning, autonomous agents, algorithmic pricing, automated procurement and autonomous digital platforms is changing this assumption.

An autonomous enterprise may be understood as an enterprise in which important commercial decisions are substantially delegated to software, algorithms, artificial intelligence or AI agents. Such systems may:

set or recommend prices;

adjust prices automatically;

select suppliers;

allocate inventory;

determine discounts;

rank products;

negotiate or recommend contractual terms;

identify competitors;

determine advertising strategies;

make investment or production decisions;

personalize offers;

optimize supply chains; and

interact with other automated systems.

Competition law therefore faces an important question:

Can existing competition-law principles effectively regulate anti-competitive conduct when the immediate decision-maker is an autonomous algorithm rather than a human executive?

The basic principle is that automation does not, by itself, remove conduct from competition law. Competition authorities increasingly examine whether algorithms facilitate collusion, exclusion, self-preferencing, information exchange, market foreclosure or other anti-competitive effects. The OECD has specifically identified both the efficiency benefits and competition risks associated with algorithmic decision-making.

2. Meaning of Autonomous Enterprises

An autonomous enterprise is not necessarily a completely human-free business.

It is better understood as a business in which commercial decision-making has been delegated to automated systems to a significant degree.

Examples

A. Autonomous pricing

An AI system automatically determines prices according to:

demand;

competitor prices;

inventory;

consumer behaviour;

time;

location; and

predicted willingness to pay.

B. Autonomous procurement

Software may automatically select suppliers and determine purchasing quantities.

C. Autonomous distribution

An AI system may determine:

which seller receives visibility;

which product is recommended;

which customer receives an offer; and

which seller obtains access to logistics facilities.

D. Autonomous negotiation

AI agents may communicate with other firms' systems and negotiate:

prices;

delivery terms;

discounts;

supply;

advertising;

contractual conditions.

E. Autonomous platform governance

A digital platform may use algorithms to determine:

ranking;

access;

commissions;

search visibility;

recommendation;

account restrictions;

seller eligibility.

Thus, autonomy can exist at several levels rather than being an all-or-nothing condition.

3. Why Autonomous Enterprises Create New Competition-Law Problems

Traditional competition law generally focuses on:

agreements;

concerted practices;

unilateral conduct by dominant firms;

mergers and acquisitions;

market power;

exclusionary conduct; and

consumer and competitive effects.

Autonomous enterprises complicate each of these areas.

4. Algorithmic Collusion

One of the most important issues is algorithmic collusion.

Suppose several competing businesses independently use AI pricing systems.

Each algorithm:

observes competitors;

predicts their reactions;

changes prices;

learns from market responses; and

attempts to maximize profits.

The systems may eventually produce parallel or elevated prices without direct human communication.

This raises a difficult distinction between:

Legitimate independent adaptation

A business independently observes market conditions and changes its price.

Anti-competitive coordination

Businesses deliberately use a common mechanism, information channel or arrangement to coordinate competitive behaviour.

Competition law must therefore determine whether the conduct constitutes:

an agreement;

concerted practice;

exchange of competitively sensitive information;

hub-and-spoke coordination;

conscious coordination; or

unilateral conduct.

The CMA has expressly recognized that pricing algorithms can produce efficiencies but may also facilitate price fixing or coordination, particularly where competing firms use a common system.

5. Autonomous Agents and the Concept of Agreement

Traditional cartel law often asks:

Did the competitors communicate and agree?

With autonomous enterprises, the question may become:

Did the businesses intentionally design, configure or use autonomous systems in a manner that substitutes automated coordination for direct human communication?

This creates several possibilities.

Model 1 – Human-to-human agreement

Managers agree to fix prices and use software to implement the agreement.

This is relatively straightforward.

Model 2 – Human agreement + autonomous execution

Managers agree on a competitive strategy and an AI system automatically implements it.

The existence of the algorithm does not eliminate liability.

Model 3 – Common algorithm

Several competitors use the same pricing provider.

The provider receives sensitive information from competing businesses and generates pricing recommendations.

This creates substantial information-exchange and coordination concerns.

Model 4 – Pure autonomous interaction

Two independent AI agents independently learn that higher prices are profitable and begin responding to each other.

This presents a much harder legal question because traditional notions of human communication and intention may be absent.

6. Information Exchange Through Autonomous Systems

Competition law generally treats the exchange of competitively sensitive information as potentially problematic.

Important information may include:

future prices;

discounts;

output;

inventory;

costs;

capacity;

customer information;

business strategy;

future production plans.

An autonomous system can collect and process enormous amounts of such information.

Consequently, the competition authority may need to examine:

who supplied the information;

whether the information was confidential;

who could access it;

how the algorithm processed it;

whether competitors knew about the process;

whether the information influenced commercial decisions; and

whether competition was reduced.

7. Self-Learning Algorithms

Machine-learning systems may behave differently from conventional software.

A conventional program generally follows predetermined instructions.

A machine-learning system may:

receive data;

identify patterns;

generate predictions;

receive feedback;

change its behaviour; and

repeat the process.

This creates a problem of explainability.

A firm may argue:

“We did not instruct the algorithm to coordinate prices.”

The competition authority may nevertheless need to determine:

how the algorithm was trained;

what objectives were programmed;

what data it received;

what constraints were imposed;

whether the firm monitored its behaviour;

whether the firm benefited from the resulting conduct; and

whether corrective controls existed.

Therefore, competition compliance may increasingly require algorithmic governance, not merely traditional legal policies.

8. Autonomous Enterprises and Abuse of Dominance

Autonomous enterprises can create problems under abuse-of-dominance rules.

A dominant AI-driven enterprise could potentially use its system to:

discriminate against competitors;

favour its own products;

reduce rival visibility;

restrict access to data;

impose discriminatory terms;

exclude competing suppliers;

tie products;

bundle services;

impose loyalty mechanisms;

manipulate ranking; or

prevent interoperability.

The central legal question remains:

Does the conduct of the dominant enterprise restrict competition without sufficient objective justification?

Automation changes the mechanism, but not necessarily the underlying competition-law analysis.

9. Algorithmic Self-Preferencing

A platform controlling a marketplace may simultaneously act as:

platform operator;

intermediary;

seller;

advertiser; and

data processor.

Its autonomous ranking algorithm may favour the platform's own products.

For example:

Platform → collects market data → identifies successful third-party products → launches competing product → algorithm gives platform product better visibility.

This can potentially create:

data advantages;

ranking advantages;

distribution advantages;

scale advantages; and

foreclosure of competitors.

The Google Shopping litigation is an important example of competition-law scrutiny of preferential treatment of a dominant platform's own specialised service. The EU General Court characterized the conduct as favouring Google's own specialised search results and upheld the Commission's infringement finding, subject to the subsequent judicial history.

10. Autonomous Enterprises and Data Concentration

AI systems require large quantities of data.

Data may therefore become a strategic competitive asset.

An autonomous enterprise with access to extensive data may obtain advantages in:

prediction;

pricing;

personalization;

fraud detection;

advertising;

product development;

logistics;

demand forecasting.

The competition concern becomes greater when the dominant enterprise controls data that rivals cannot realistically reproduce.

This may produce a data feedback loop:

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

Such feedback may strengthen network effects and entry barriers.

11. Network Effects

Autonomous digital businesses may benefit from strong network effects.

The value of a platform can increase as more:

consumers join;

sellers participate;

advertisers use it;

developers build applications;

data is generated.

AI can strengthen this effect because more users generate more training data.

This may result in:

Scale → Data → Better AI → Better service → More users → More scale.

Competition law must therefore consider dynamic market power, not merely current prices.

12. Autonomous Enterprises and Predatory Pricing

AI may enable extremely sophisticated pricing strategies.

A dominant firm may theoretically use an algorithm to:

identify vulnerable competitors;

temporarily lower prices;

target particular geographic areas;

offer personalized discounts;

increase prices after rivals exit.

Competition authorities may therefore need to examine:

cost benchmarks;

duration of low pricing;

targeted nature of pricing;

recoupment;

exclusionary intent or effect;

consumer benefits; and

long-term competitive effects.

Automation does not create an exemption from predatory-pricing rules.

13. Personalized Pricing

AI allows businesses to estimate individual willingness to pay.

This may create:

individualized discounts;

individualized prices;

targeted promotions;

dynamic pricing.

Personalization is not automatically anti-competitive.

However, competition concerns may arise if personalized pricing is used by a dominant enterprise to:

discriminate against rivals' customers;

selectively target competing firms;

prevent switching;

exploit lock-in;

reward exclusivity; or

make entry more difficult.

Competition authorities may therefore have to distinguish legitimate price optimization from exclusionary or exploitative strategies.

14. Autonomous Enterprises and Vertical Foreclosure

An autonomous enterprise may control several levels of the supply chain.

For example:

AI manufacturer → marketplace → logistics → payment → advertising → consumer interface

The enterprise can potentially use information from one level to strengthen its position at another.

Possible practices include:

tying;

bundling;

exclusive dealing;

discriminatory access;

preferential ranking;

refusal to supply;

loyalty rebates;

discriminatory commissions.

This makes traditional vertical competition-law analysis increasingly important in AI-driven markets.

15. Autonomous Enterprises and Essential Facilities

Some AI-driven businesses may control infrastructure that competitors need.

Examples could include:

dominant cloud infrastructure;

app stores;

digital marketplaces;

payment infrastructure;

essential datasets;

technical interfaces;

interoperability tools.

The essential facilities doctrine may become relevant where a dominant enterprise controls an indispensable facility.

The traditional European approach, particularly Bronner, requires stringent conditions before a refusal to provide access can amount to abuse.

The development of digital ecosystems has nevertheless raised questions about whether traditional essential-facilities principles can adequately address platform interoperability.

16. Autonomous Enterprises and Interoperability

Competition can suffer if an autonomous enterprise deliberately prevents its AI system from communicating with rival systems.

Examples include:

blocking APIs;

restricting data portability;

preventing interoperability;

refusing technical access;

limiting third-party applications;

preventing competing AI agents from interacting with a platform.

Interoperability can therefore become a competition parameter.

The law may need to distinguish between:

Legitimate technological protection

Security, privacy and intellectual-property protection.

Exclusionary interoperability restrictions

Restrictions primarily designed to prevent competitors from entering or expanding.

17. Autonomous Enterprises and Merger Control

Autonomous enterprises also create challenges for merger law.

A large enterprise may acquire:

an AI start-up;

a data provider;

an algorithm developer;

a cloud technology company;

a robotics company;

a specialised AI model.

The acquired company may have:

little current revenue;

few employees;

substantial technological capability;

valuable data;

significant future competitive potential.

Traditional turnover thresholds may therefore fail to capture some strategically important acquisitions.

Competition authorities increasingly need to consider:

innovation;

potential competition;

data;

intellectual property;

technology;

future markets;

ecosystem effects.

This is particularly important where the target could become a future competitor.

18. Autonomous Enterprises and Killer Acquisitions

A dominant autonomous platform may acquire a small AI company before it becomes a significant competitor.

The acquisition may eliminate:

future innovation;

a technological challenger;

an alternative AI architecture;

a new distribution model.

Therefore, merger control may need to examine not only:

“Does the target currently compete?”

but also:

“Could the target become an important competitive constraint?”

19. Autonomous Enterprises and Consumer Choice

AI recommendations can influence what consumers see.

A recommendation system may determine:

which products appear first;

which seller is recommended;

which advertisement is shown;

which search result is promoted;

which product is excluded.

Consequently, competition can shift from price competition to visibility competition.

A rival may technically remain in the market but become commercially invisible.

This creates an important modern competition concept:

“Competition for algorithmic visibility.”

20. Autonomous Enterprises and Market Definition

Traditional market definition often examines:

products;

services;

geographic boundaries;

substitutability;

demand;

supply.

Autonomous enterprises may complicate this because one AI ecosystem can operate across several interconnected markets.

For example:

Search → advertising → data → cloud → AI assistant → marketplace

The competition authority may need to consider:

multi-sided markets;

ecosystem effects;

complementary products;

zero-price services;

data;

switching costs;

interoperability.

21. Autonomous Enterprises and Dynamic Competition

Traditional competition analysis can sometimes emphasize current market shares.

Autonomous enterprises require greater attention to:

innovation;

technological development;

future competition;

research and development;

data accumulation;

learning effects;

network effects.

An enterprise with a relatively small current market share may possess strategically important AI technology capable of becoming a significant competitive constraint.

22. Case Law

Case 1: Eturas UAB and Others v Lietuvos Respublikos konkurencijos taryba

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

This is one of the most relevant European cases concerning automated online systems.

Several travel agencies used a common computerized booking system.

The system administrator sent a message concerning restrictions on discounts, and the system automatically limited the discounts available to customers.

The CJEU considered whether conduct implemented through a common computerized system could constitute a concerted practice.

Principle

Technology does not prevent competition law from applying to coordinated conduct.

The case demonstrates that courts can examine:

digital communications;

automated implementation;

common systems;

tacit coordination; and

evidence of participation.

It is highly relevant to autonomous enterprises because the anti-competitive conduct may be implemented automatically rather than manually.

Case 2: T-Mobile Netherlands BV v Raad van bestuur van de Nederlandse Mededingingsautoriteit

Case C-8/08

This case concerned the concept of concerted practice.

The CJEU held that, in appropriate circumstances, even a single meeting can be sufficient to establish a concerted practice where the legal requirements are satisfied.

Importance for autonomous enterprises

The case establishes that competition law does not necessarily require prolonged or formal coordination.

In an AI environment, a competition authority may therefore examine:

the nature of communication;

the information exchanged;

the firms' subsequent conduct; and

the competitive significance of the coordination.

The case is relevant to the broader question of how traditional concerted-practice principles adapt to increasingly automated markets.

Case 3: United States v David Topkins

United States v David Topkins involved an online price-fixing scheme concerning posters sold through an online marketplace.

The U.S. Department of Justice described it as its first criminal prosecution specifically targeting price fixing in an e-commerce marketplace. Topkins pleaded guilty under an agreement involving the use of pricing algorithms.

Principle

The use of software does not transform illegal price fixing into lawful competition.

The case demonstrates:

Human agreement + algorithmic implementation = potential conventional antitrust liability.

Importance

For autonomous enterprises, companies cannot simply argue:

“The computer fixed the price, not the employees.”

Legal responsibility may still attach to the enterprise and individuals depending on the facts.

Case 4: Samir Agrawal v Competition Commission of India

This Indian litigation concerned allegations that Ola and Uber's algorithmic pricing mechanisms facilitated price fixing between drivers.

The allegation was that drivers were effectively required to accept fares calculated through the platforms' algorithms.

The CCI initially found no prima facie agreement establishing the alleged contravention and closed the matter. The appellate litigation ultimately reached the Supreme Court, which upheld the closure of the matter.

Importance

This is particularly important for Indian competition law.

It demonstrates that:

Algorithmic pricing alone does not automatically establish a cartel.

The authority must establish the necessary legal elements of an agreement, arrangement or concerted practice.

This is a useful safeguard against treating every parallel algorithmic pricing outcome as unlawful coordination.

Case 5: Google Shopping

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

Google's treatment of its own comparison-shopping service in search results was examined under Article 102 TFEU.

The General Court addressed Google's preferential display of its own specialised search results and the effect on competing services.

Importance for autonomous enterprises

AI-driven ranking systems can make millions of commercial decisions automatically.

A dominant platform may therefore embed competitive preferences into:

ranking algorithms;

recommendation systems;

search systems;

advertising algorithms.

The case demonstrates that automated ranking does not place conduct outside abuse-of-dominance rules.

Case 6: Amazon Marketplace

The European Commission investigated Amazon's use of non-public marketplace seller data.

The Commission's preliminary concerns included the alleged use of third-party seller data by Amazon's retail business for decisions concerning products, suppliers, inventory and pricing.

The UK CMA separately accepted commitments addressing concerns relating to Amazon's use of seller data, Buy Box selection and Prime-related practices. The investigation was closed after the commitments decision.

Importance

The case demonstrates the interaction between:

data + algorithms + platform power + vertical integration.

An autonomous enterprise may continuously collect market information and use it to improve its own competing business.

This can create a competitive feedback loop:

Third-party activity → data collection → AI analysis → improved own products → stronger platform position.

Case 7: United States v Microsoft Corp.

253 F.3d 34 (D.C. Cir. 2001)

Microsoft's conduct concerning Internet Explorer and the Windows operating-system ecosystem was examined under U.S. antitrust law.

Importance for autonomous enterprises

Although this case predates modern AI, its principles remain relevant.

It demonstrates that competition law can address the strategic use of technological architecture to:

protect a dominant position;

disadvantage rivals;

control distribution channels;

extend monopoly power into related markets.

Modern autonomous enterprises can potentially reproduce similar strategies through algorithms rather than traditional software architecture.

Case 8: RealPage Algorithmic Pricing Litigation

The U.S. Department of Justice sued RealPage, alleging that its revenue-management software enabled competing landlords to share competitively sensitive information and align rental pricing.

The DOJ alleged violations of Sections 1 and 2 of the Sherman Act.

The later proceedings produced proposed and final settlements concerning various participants; as of 2026, the DOJ's case materials show continuing proceedings involving several landlord defendants and a proposed judgment concerning RealPage itself.

Importance

RealPage is particularly significant because the alleged mechanism is close to the autonomous-enterprise problem:

Competitor data → algorithm → pricing recommendation → automated or highly influential pricing decisions.

The case illustrates why competition authorities may examine not merely the final price, but also:

the data supplied to the algorithm;

the algorithm's design;

pricing recommendations;

monitoring mechanisms;

user incentives; and

the degree of human discretion.

23. What These Cases Teach About Autonomous Enterprises

The cases collectively demonstrate several principles.

24. Adaptation of Competition Law

Competition law does not necessarily need to be completely rewritten.

Instead, it may need to be adapted in several ways.

A. Adaptation of the Concept of Agreement

Competition authorities should examine whether:

algorithms are jointly supplied;

competitors intentionally use the same pricing mechanism;

sensitive information is shared;

firms knowingly rely upon coordinated systems;

software is configured to facilitate coordination.

The focus should remain on legally relevant conduct rather than merely on the existence of AI.

25. Algorithmic Audit

Competition authorities may increasingly need the power to examine:

source code;

model architecture;

training data;

instructions;

system objectives;

logs;

model outputs;

API interactions;

pricing histories.

This may become the digital equivalent of examining emails and internal documents in conventional cartel investigations.

26. Algorithmic Compliance Programmes

Companies using autonomous systems should establish competition-compliance controls.

These may include:

competition-law review before deployment;

restrictions on competitor data;

monitoring of pricing algorithms;

documentation of model objectives;

independent algorithm audits;

human review of high-risk decisions;

automatic compliance alerts;

restrictions on common pricing systems;

records of algorithmic changes; and

procedures for correcting anti-competitive outcomes.

27. Human Oversight

Human oversight is especially important where an autonomous system has significant market power.

A company should not simply state:

“The AI made the decision.”

The relevant questions are:

Who designed the system?

Who selected the objective?

Who supplied the data?

Who approved deployment?

Who monitored the output?

Who benefited?

Who could stop the system?

This creates a concept of responsible autonomy.

28. Transparency and Explainability

Competition authorities may require greater explanation of algorithmic decisions.

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

A more proportionate approach may involve:

audit trails;

decision logs;

model documentation;

testing;

explanations of ranking criteria;

records of changes;

access to independent technical experts.

The objective should be to determine whether the system produces anti-competitive effects without unnecessarily destroying legitimate trade secrets.

29. Competition Law and AI Agents Negotiating With Each Other

A future market may involve:

AI Agent A → AI Agent B → AI Agent C

Agents could negotiate:

prices;

supply;

transportation;

advertising;

procurement.

This creates a new legal problem:

Who is responsible for the AI agent's conduct?

Possible approaches include:

responsibility of the enterprise deploying the agent;

responsibility of the enterprise controlling the agent;

responsibility of the software provider where legally appropriate;

shared responsibility depending on the facts.

Competition law should focus on human and corporate control over autonomous systems, rather than treating the AI itself as an independent legal person.

30. Autonomous Enterprises and Indian Competition Law

In India, the principal statutory framework remains the Competition Act, 2002.

Important provisions include:

Section 3

Deals with anti-competitive agreements.

This can become relevant to:

algorithmic price fixing;

information exchange;

hub-and-spoke arrangements;

common software systems.

Section 4

Deals with abuse of dominant position.

Potential autonomous-enterprise issues include:

self-preferencing;

discriminatory access;

tying;

refusal to deal;

exclusionary algorithms;

leveraging;

data-based foreclosure.

Sections 5 and 6

Concern combinations and merger control.

They become important where dominant enterprises acquire:

AI start-ups;

data companies;

algorithm providers;

emerging technological competitors.

Section 19

Provides the framework for inquiry by the Competition Commission of India.

In digital markets, investigation may require technical and economic analysis of:

algorithms;

data;

network effects;

switching costs;

platform architecture.

31. New Evidence Problems

Autonomous enterprises create enormous quantities of electronic evidence.

Important evidence may include:

algorithm logs;

training datasets;

API records;

model versions;

system instructions;

pricing histories;

internal communications;

server records;

automated decision records.

Competition authorities therefore need technical capabilities to reconstruct:

What the algorithm knew → what it was instructed to do → what it actually did → what the enterprise knew → what competitive effect followed.

32. The Problem of Black-Box Algorithms

A major problem is the black-box effect.

The firm may know the input and output but have difficulty explaining exactly how the model reached the decision.

This creates evidentiary challenges.

Competition authorities may therefore increasingly rely on:

statistical analysis;

controlled testing;

algorithmic auditing;

expert evidence;

source-code analysis;

data analysis;

economic modelling.

33. Autonomous Enterprises and Consumer Welfare

Autonomous systems can generate significant benefits.

Potential benefits

lower operating costs;

faster decision-making;

better inventory management;

lower transaction costs;

improved forecasting;

personalized products;

increased innovation;

better resource allocation.

Therefore, competition law should not assume:

AI = anti-competitive.

The correct approach is to distinguish efficiency-enhancing automation from automation that restricts competition.

The OECD similarly recognizes that algorithms may produce significant efficiency-enhancing and pro-competitive effects while also creating competition risks.

34. Potential Harms

Autonomous enterprises may nevertheless create:

1. Higher prices

Through algorithmic coordination.

2. Reduced choice

Through automated exclusion.

3. Less innovation

When dominant platforms suppress emerging competitors.

4. Entry barriers

Through data and network effects.

5. Market concentration

Through scale and learning advantages.

6. Discriminatory access

Through algorithmic ranking.

7. Consumer lock-in

Through ecosystem integration.

8. Information asymmetry

Where consumers cannot understand automated decisions.

35. Need for a New Competition-Law Framework?

A complete new competition statute may not always be necessary.

Existing principles can often be applied to new technology.

However, adaptation may be required in:

evidence;

market definition;

algorithmic audits;

merger review;

data analysis;

technical expertise;

remedies;

compliance;

interoperability.

Thus, the better approach can be described as:

Traditional competition principles + technological adaptation + stronger investigative capability.

36. Possible Remedies

Competition authorities may use several remedies.

A. Behavioural remedies

prohibit specific algorithmic practices;

require non-discriminatory ranking;

restrict data use;

require access;

prohibit certain information exchanges.

B. Structural remedies

In exceptional circumstances:

separation of business units;

divestiture;

restrictions on acquisitions.

C. Technical remedies

algorithmic audits;

independent monitoring;

interoperability;

data portability;

API access.

D. Governance remedies

compliance officers;

algorithmic risk committees;

reporting obligations;

audit trails;

human oversight.

37. Challenges for Competition Authorities

1. Technical complexity

Competition authorities require AI and data-science expertise.

2. Speed

AI markets can change much faster than traditional investigations.

3. Evidence

The most important evidence may exist inside technical systems.

4. Causation

It may be difficult to prove that an algorithm caused a particular competitive harm.

5. Intent

Autonomous systems complicate the traditional concept of human intention.

6. False positives

Parallel algorithmic outcomes do not necessarily prove collusion.

7. Innovation

Over-regulation may discourage beneficial AI development.

8. Cross-border enforcement

Autonomous systems can operate across multiple jurisdictions simultaneously.

38. Future Direction of Competition Law

Competition law relating to autonomous enterprises is likely to develop around several principles:

Principle 1

Automation is not an exemption from competition law.

Principle 2

Algorithms should be assessed according to their competitive function and effects.

Principle 3

The use of AI should not automatically be treated as evidence of collusion.

Principle 4

Human control, design and deployment remain legally important.

Principle 5

Data and network effects should be considered in assessing market power.

Principle 6

Dynamic innovation competition must be considered alongside current market shares.

Principle 7

Merger control should consider potential technological competition.

Principle 8

Competition authorities require technical investigative capabilities.

39. Key Distinction: Autonomous Conduct vs Autonomous Coordination

This distinction is extremely important for examination.

Autonomous conduct

One enterprise independently uses AI to optimize its own business.

Generally: not inherently anti-competitive.

Autonomous coordination

Several competitors use systems that intentionally or knowingly coordinate their competitive behaviour.

Potential issue: cartel/concerted-practice liability.

Autonomous exclusion

A dominant enterprise's algorithm systematically disadvantages rivals.

Potential issue: abuse of dominance.

Autonomous acquisition

A dominant enterprise acquires an emerging AI competitor.

Potential issue: merger control and loss of potential competition.

40. Overall Legal Analysis

When examining an autonomous enterprise, a competition authority should ask:

Step 1 – What is the relevant market?

Identify:

product/service;

geographic market;

multi-sided nature;

digital ecosystem.

Step 2 – What degree of autonomy exists?

Determine whether the system:

recommends;

decides;

executes;

learns;

negotiates.

Step 3 – Who controls the system?

Examine:

enterprise;

management;

programmers;

third-party provider.

Step 4 – What data does it use?

Determine whether it uses:

public data;

private data;

competitor data;

customer data;

market-sensitive information.

Step 5 – What conduct results?

Consider:

price fixing;

exclusion;

self-preferencing;

tying;

discrimination;

refusal to deal;

information exchange.

Step 6 – What is the competitive effect?

Examine:

price;

output;

quality;

choice;

innovation;

entry;

market structure.

Step 7 – Are there efficiencies?

Consider:

cost savings;

innovation;

improved service;

better allocation.

Step 8 – What remedy is proportionate?

Possible remedies include:

behavioural;

structural;

technical;

interoperability;

monitoring;

algorithmic compliance.

41. Conclusion

Autonomous enterprises represent an important development in modern competition law.

The central transformation is from:

Human decision-making → automated decision-making → increasingly autonomous commercial decision-making.

This does not make traditional competition law irrelevant.

Instead, it requires competition law to adapt its application to:

algorithms;

artificial intelligence;

autonomous agents;

data;

network effects;

digital ecosystems;

algorithmic pricing;

automated ranking;

autonomous negotiation;

technological acquisitions.

Cases such as Eturas, T-Mobile Netherlands, Topkins, Samir Agrawal, Google Shopping, Amazon Marketplace, Microsoft and RealPage demonstrate that competition law is already dealing with different parts of this technological transformation.

The central legal principle can therefore be stated as:

An enterprise cannot escape competition law merely because its commercial decisions are made or implemented by an algorithm. At the same time, autonomous behaviour should not automatically be treated as unlawful collusion. Liability must depend upon the applicable competition-law elements, the enterprise's conduct and control, and the actual or likely effects on competition.

42. Short Exam Revision

Autonomous Enterprise

Enterprise that delegates substantial commercial decision-making to AI, algorithms or automated systems.

Major Competition Issues

Algorithmic price fixing

Autonomous collusion

Information exchange

Self-preferencing

Data concentration

Network effects

Predatory pricing

Personalized pricing

Vertical foreclosure

Interoperability restrictions

Killer acquisitions

Innovation suppression

Market concentration

Consumer lock-in

Algorithmic discrimination

Important Cases

Eturas UAB v Lithuanian Competition Council (C-74/14) – computerized system and concerted practice.

T-Mobile Netherlands (C-8/08) – concerted practices.

United States v David Topkins – online algorithmic price fixing.

Samir Agrawal v CCI – Ola/Uber algorithmic pricing and Indian Section 3 analysis.

Google Shopping (T-612/17) – preferential treatment/self-preferencing.

Amazon Marketplace – platform data and competitive concerns.

United States v Microsoft – technological exclusion and platform power.

RealPage litigation – algorithmic pricing and competitor information.

Exam Formula

Autonomous enterprise → AI/algorithm → data → network effects → market power → algorithmic conduct → collusion/exclusion/self-preferencing → competitive effects → efficiencies → competition-law liability → technical/behavioural/structural remedies.