Competition Law and Strategic Competition Oversight in Self-Organizing Markets

1. Introduction

Self-organizing markets are markets in which firms, consumers, platforms, algorithms, or artificial-intelligence systems continuously adjust their conduct in response to market signals without requiring a central coordinator. Examples include algorithmic pricing markets, online marketplaces, app ecosystems, digital advertising exchanges, autonomous trading systems, ride-hailing platforms, hotel-booking platforms, and AI-driven recommendation systems.

The central competition-law problem is that competitive outcomes may emerge from decentralized technological processes. A market may therefore exhibit coordinated pricing, exclusion, discrimination, or concentration even where there is no traditional cartel meeting or explicit written agreement.

The OECD has specifically recognized that algorithms can facilitate both coordinated conduct and unilateral exclusionary or exploitative conduct, while noting that traditional competition-law concepts remain relevant.

Strategic competition oversight therefore asks:

How should competition authorities supervise markets in which competitive conditions are continuously shaped by algorithms, platforms, data, automated decision-making, and self-learning systems?

2. Meaning of Self-Organizing Markets

A self-organizing market generally possesses several characteristics:

1. Decentralized decision-making – firms make decisions independently.
2. Continuous adaptation – algorithms react rapidly to competitors and consumers.
3. Data-driven optimization – pricing, ranking, advertising and recommendations depend on large datasets.
4. Network effects – more users attract more users and sellers.
5. Feedback loops – an algorithm's previous decisions influence future decisions.
6. Low-cost automated coordination – software can rapidly respond to competitors' conduct.
7. Platform dependence – businesses may depend upon a small number of digital infrastructures.
8. Dynamic market boundaries – products and services may change faster than conventional market-definition exercises.

The OECD has emphasized that AI and algorithmic systems can lower entry barriers in some circumstances while simultaneously creating risks concerning data access, model restrictions, coordination and contestability.

3. Why Traditional Competition Oversight Becomes Difficult

Traditional competition law often assumes identifiable human decision-makers.

In a self-organizing market, however:

Human decision → Algorithm → Market reaction → Algorithmic learning → New decision → Market reaction

may occur thousands or millions of times.

This creates several problems.

A. Absence of an obvious agreement

Two competitors may independently deploy algorithms that observe one another's prices and automatically respond.

The resulting prices may become highly coordinated without a conventional agreement.

The legal question becomes whether the outcome resulted from:

• independent adaptation;
• exchange of competitively sensitive information;
• a common algorithm;
• a third-party intermediary;
• explicit coordination;
• or a combination of these factors.

The OECD has identified precisely this distinction between explicit algorithmic coordination and merely parallel outcomes generated by independently operating algorithms.

B. Speed of market evolution

A conventional investigation may take years, while an algorithm can modify competitive conditions in seconds.

C. Opacity

Authorities may not know:

• what data the algorithm receives;
• what variables it optimizes;
• how it ranks competitors;
• whether it learns from competitors' conduct;
• whether humans can override it;
• or whether the algorithm has developed unexpected strategies.

D. Feedback loops

An algorithm may initially behave competitively but gradually learn that coordinated pricing is more profitable.

This creates a difficult distinction between intentional coordination and autonomous convergence.

4. Strategic Competition Oversight

Strategic oversight goes beyond investigating completed violations.

It involves continuous monitoring of competitive conditions.

A useful framework is:

Market monitoring → Risk identification → Algorithmic audit → Competitive assessment → Early intervention → Remedy → Continuous review

This approach is increasingly relevant because the OECD has observed that competition authorities are investing in algorithmic auditing, technical expertise and methods for detecting digital competitive harm.

5. Major Competition Concerns

A. Algorithmic Collusion

Algorithms can facilitate:

• price fixing;
• market allocation;
• output restrictions;
• coordination of discounts;
• information exchange;
• monitoring of rivals.

The important distinction is between:

Explicit algorithmic collusion

Competitors communicate and deliberately program algorithms to implement an agreement.

Tacit algorithmic coordination

Independent algorithms observe market conditions and autonomously converge toward similar strategies.

The first is much more comfortably addressed through existing cartel rules. The second raises difficult questions concerning proof and the limits of competition law.

6. Self-Learning Algorithms

Self-learning algorithms present a particularly important challenge.

A conventional algorithm might be programmed:

"If competitor raises price, increase our price by 5%."

A machine-learning system might instead learn:

"Maintaining a certain price relationship with competitors maximizes long-term expected profit."

The second system may generate coordination without a human expressly programming the prohibited result.

This raises the question:

Who is legally responsible for an anticompetitive outcome produced by autonomous learning?

Potentially relevant actors include:

• the firm using the system;
• the algorithm developer;
• the platform hosting it;
• a common software supplier;
• or several firms collectively.

Current competition law generally focuses on the conduct and economic relationship of undertakings rather than treating autonomous algorithmic behaviour as an independent legal actor.

7. Platform Governance and Self-Organizing Markets

Large platforms can effectively become market organizers.

They may determine:

• search rankings;
• product visibility;
• access conditions;
• commission structures;
• advertising placement;
• seller eligibility;
• recommendation systems;
• interoperability;
• data access;
• and pricing architecture.

Consequently, a platform can simultaneously be:

market participant + infrastructure provider + rule-maker + data intermediary + algorithmic gatekeeper.

This creates a structural competition concern.

The European Commission's recent DMA enforcement illustrates the move toward ex ante oversight: in July 2026, the Commission announced findings against Google concerning self-preferencing in Search and restrictions on steering users toward alternative purchasing channels.

8. Data as a Competitive Infrastructure

Self-organizing digital markets frequently depend on data.

A dominant undertaking may control:

• consumer behaviour data;
• transaction data;
• search data;
• location data;
• seller-performance data;
• advertising data;
• technical telemetry;
• AI-training data.

Data concentration may produce competitive advantages through a feedback loop:

More users → More data → Better algorithm → Better service → More users → More data

This can create data-driven market power even where conventional barriers to entry appear modest.

9. Interoperability as a Competition Remedy

Where market power depends on ecosystem control, interoperability can become an important remedy.

Possible obligations include:

• API access;
• data portability;
• interoperability with competing services;
• non-discriminatory access;
• technical compatibility;
• access to essential datasets.

The EU's 2026 DMA measures concerning Google's Android ecosystem and search data illustrate this approach. The Commission required measures intended to provide competing AI services with access to relevant Android functionality and third-party search engines with access to certain search data.

10. Algorithmic Self-Preferencing

A platform's algorithm may systematically favour its own products.

For example:

Platform → controls ranking → operates competing service → ranking algorithm → own service receives preferential visibility

The concern is not merely that the platform has a large market share. The issue is whether control of the algorithmic infrastructure is being used to disadvantage rivals.

This has become a major area of digital competition enforcement.

11. Price Discrimination

Self-organizing systems can automatically vary prices according to:

• location;
• browsing history;
• purchasing behaviour;
• device;
• time;
• demand;
• customer profile;
• willingness to pay.

Personalized pricing is not automatically unlawful.

Competition concerns arise where algorithmic pricing is used to:

• exploit market power;
• exclude rivals;
• discriminate in competitively significant ways;
• facilitate coordination;
• or reinforce dominance.

12. Essential-Facility-Like Digital Infrastructure

Some digital ecosystems can acquire characteristics resembling essential facilities.

Examples may include:

• dominant app stores;
• payment networks;
• cloud infrastructure;
• major search indexes;
• digital identity infrastructure;
• interoperability interfaces;
• dominant marketplace infrastructure.

Competition oversight may therefore examine whether access is:

• denied;
• delayed;
• technically degraded;
• excessively priced;
• discriminatory;
• conditioned on unrelated services.

13. Six Important Case Laws / Enforcement Decisions

1. Eturas UAB and Others v Lietuvos Respublikos konkurencijos taryba, Case C-74/14 (CJEU, 2016)

This is one of the most important European cases involving automated digital coordination.

Travel agencies used a common electronic booking system. The system administrator communicated a message concerning a limitation on discounts, and the system was technically modified to implement the restriction.

The CJEU held that, where the relevant conditions were satisfied, knowledge of the message and continued participation could support a presumption of participation in a concerted practice, although the presumption was rebuttable.

Principle

Digital implementation does not prevent Article 101 TFEU from applying.

Importance

Eturas demonstrates that competition law can examine:

communication + technological implementation + participant knowledge + subsequent conduct.

It is therefore highly relevant to self-organizing markets.

2. United States v. David Topkins (2015)

The U.S. Department of Justice prosecuted an online poster-pricing conspiracy involving Amazon Marketplace.

The participants agreed to fix prices and implemented the agreement through pricing algorithms. The indictment described algorithms designed to coordinate price changes and maintain collusive prices.

Principle

An algorithm does not immunize otherwise unlawful price fixing.

Importance

Topkins illustrates the relatively straightforward situation of:

human agreement → algorithmic implementation → coordinated prices.

It is therefore different from purely autonomous algorithmic convergence.

3. Trod Ltd and GB eye Ltd — CMA (UK, 2016)

The UK Competition and Markets Authority found that two competing online sellers agreed not to undercut each other's prices for certain products sold on Amazon's UK website.

The cartel was implemented using automated repricing software. The CMA imposed a fine on Trod, while GB eye received leniency after reporting the conduct.

Principle

Automated implementation is no defence to cartel liability.

Importance

The case demonstrates the distinction between:

illegal human coordination + automated execution

and genuinely independent algorithmic behaviour.

It also demonstrates why authorities must examine the configuration and purpose of pricing software, not merely observed prices.

4. HRS Hotel Reservation Services — Bundeskartellamt / Düsseldorf Higher Regional Court (2013–2015)

HRS required hotels to offer its platform the lowest prices and conditions available through competing channels.

The Bundeskartellamt prohibited the best-price clauses, and the Düsseldorf Higher Regional Court confirmed the prohibition in 2015.

Principle

Platform contractual restrictions can reduce competition between platforms and restrict the pricing freedom of participating businesses.

Importance for self-organizing markets

Platform rules can shape the behaviour of an entire ecosystem.

The platform does not need to dictate every individual price directly. Contractual architecture can influence thousands of independent participants.

5. Booking.com — Bundeskartellamt / German Federal Court of Justice

The Booking.com proceedings concerned narrow price-parity clauses.

The Bundeskartellamt initially prohibited the clauses and later litigation ultimately resulted in judicial confirmation of the prohibition; the authority states that the Federal Court of Justice confirmed the prohibition in the Booking case.

Principle

A platform's contractual architecture may affect:

• platform competition;
• hotel pricing;
• entry by competing platforms;
• consumer choice.

Importance

Booking.com demonstrates that competition oversight can target ecosystem rules, rather than merely conventional pricing conduct.

6. Google Shopping — European Commission

The Google Shopping decision concerned Google's use of its dominant general search position in connection with comparison-shopping services.

The case is significant for the principle that control over a major digital infrastructure and ranking mechanism can be used in a manner that disadvantages competing services.

It illustrates an important model of modern competition oversight:

dominant infrastructure + algorithmic ranking + downstream competition = potential exclusionary conduct.

The broader OECD literature identifies self-preferencing as one of the principal forms of algorithmic unilateral conduct considered by competition authorities.

14. Additional Relevant Modern Example: Amazon

The U.S. FTC and state attorneys general sued Amazon in 2023, alleging a range of exclusionary practices involving its online-superstore and marketplace businesses, including alleged anti-discounting mechanisms and alleged preference for Amazon's own products in search results. These are allegations in litigation, rather than findings that should be treated as established violations merely because the complaint was filed.

The case illustrates the type of conduct that strategic oversight may examine:

ranking + pricing rules + seller restrictions + fulfillment requirements + advertising + platform power.

15. Lessons From the Cases

16. Strategic Oversight Model

A modern competition authority could adopt a five-layer oversight model.

Layer 1 — Structural monitoring

Monitor:

• concentration;
• switching costs;
• network effects;
• multi-homing;
• data concentration;
• interoperability.

Layer 2 — Algorithmic monitoring

Examine:

• pricing algorithms;
• ranking algorithms;
• recommendation engines;
• allocation systems;
• automated contracting;
• AI agents.

Layer 3 — Behavioural monitoring

Look for:

• parallel pricing;
• discriminatory rankings;
• exclusionary access conditions;
• sudden coordinated changes;
• loyalty restrictions;
• self-preferencing.

Layer 4 — Technical investigation

Authorities may examine:

• source code;
• model architecture;
• training data;
• logs;
• API calls;
• model inputs;
• outputs;
• version histories;
• human overrides.

The OECD specifically identifies algorithmic auditing and technical investigative capacity as increasingly important enforcement tools.

Layer 5 — Continuous remedies

Possible remedies include:

• interoperability;
• data access;
• non-discrimination;
• algorithmic transparency;
• monitoring trustees;
• behavioural commitments;
• structural separation;
• access obligations;
• restrictions on information sharing.

17. Ex Ante and Ex Post Regulation

Ex Post Competition Law

Traditional enforcement investigates conduct after it occurs.

Examples:

• cartel investigation;
• abuse of dominance;
• exclusionary conduct;
• merger review.

Ex Ante Competition Regulation

Self-organizing markets increasingly justify preventive supervision.

Examples:

• gatekeeper obligations;
• interoperability requirements;
• data-access rules;
• algorithmic auditing;
• restrictions on self-preferencing;
• advance notification of significant ecosystem changes.

The EU Digital Markets Act is an important example of this movement toward ex ante digital-market obligations. The European Commission's 2026 measures concerning Google demonstrate how such obligations can address ecosystem-level competitive conditions rather than waiting for conventional Article 102 litigation.

18. Competition Oversight of Autonomous AI Agents

The next stage is potentially more complex.

An AI agent may be capable of:

1. observing competitors;
2. negotiating prices;
3. selecting suppliers;
4. purchasing advertising;
5. changing product terms;
6. responding to demand;
7. negotiating contracts;
8. optimizing long-term profits.

Such an agent effectively becomes an autonomous economic decision-maker operating on behalf of a firm.

This creates several competition questions:

Question 1

Can an AI agent create an antitrust violation without a human expressly instructing it to do so?

Question 2

Who is responsible for its conduct?

Question 3

Should firms be required to conduct competition-risk testing before deploying autonomous agents?

Question 4

Should high-risk market algorithms be independently audited?

Question 5

Should authorities receive access to algorithmic logs during investigations?

These questions are still developing. The OECD's recent work describes agentic AI and attribution of liability as emerging issues rather than settled areas of competition law.

19. Principle of Algorithmic Accountability

A useful regulatory principle is:

A firm should not escape competition-law responsibility merely because an economically significant decision was delegated to software.

This does not mean every algorithmic outcome is unlawful.

Instead, competition analysis should distinguish:

Efficient autonomous optimization

from

algorithmically facilitated anticompetitive conduct.

That distinction is crucial because algorithms can also generate substantial efficiencies, including better matching, lower transaction costs, improved logistics and more accurate demand forecasting. The OECD expressly recognizes both the pro-competitive and anticompetitive potential of algorithms.

20. Role of Market Studies

Competition authorities should not wait for a formal complaint where a market exhibits high algorithmic risk.

Market studies can examine:

• concentration;
• pricing software;
• common algorithm providers;
• data-sharing practices;
• platform dependency;
• interoperability;
• switching costs;
• algorithmic discrimination;
• entry barriers.

This is particularly useful because algorithmic risks may become visible before a conventional competition-law infringement can be conclusively established.

21. Evidentiary Challenges

Evidence in self-organizing markets may include:

Direct evidence

• emails;
• contracts;
• instructions;
• algorithm specifications.

Digital evidence

• source code;
• logs;
• API records;
• model versions;
• database records.

Economic evidence

• price movements;
• market shares;
• margins;
• demand elasticity;
• switching rates.

Algorithmic evidence

• model outputs;
• training data;
• optimization objectives;
• reward functions;
• decision trees;
• reinforcement-learning behaviour.

Circumstantial evidence

Repeated coordinated behaviour + common algorithm + information exchange + economically irrational independent conduct

may provide an evidentiary basis for deeper investigation, subject to the legal standard applicable in the jurisdiction.

Eturas is particularly significant because the CJEU recognized that competition infringements can be established through objective and consistent indicia, while preserving the presumption of innocence.

22. Remedies for Self-Organizing Markets

1. Transparency remedies

Require disclosure of relevant:

• ranking principles;
• pricing practices;
• access criteria.

2. Interoperability

Allow competitors to connect to critical digital infrastructure.

3. Data portability

Reduce switching costs and data-based entry barriers.

4. Non-discrimination

Require platforms to apply equivalent access conditions to competing and affiliated businesses.

5. Algorithmic audits

Independent testing can examine whether systems systematically produce exclusionary or discriminatory outcomes.

6. Information-firewall remedies

Prevent competitively sensitive data obtained from sellers or competitors from being used to advantage the platform's own competing business.

7. Structural remedies

In exceptional circumstances, separation of infrastructure and downstream commercial operations may be considered.

23. Challenges of Strategic Oversight

Strategic oversight itself creates risks.

A. False positives

Similar algorithmic behaviour does not necessarily prove collusion.

B. Innovation chilling

Excessive intervention may discourage beneficial AI and algorithmic experimentation.

C. Confidentiality

Algorithmic audits may involve highly sensitive trade secrets.

D. Technical complexity

Competition authorities require economists, lawyers, computer scientists and data specialists.

E. International enforcement

A pricing algorithm may be designed in one country, hosted in another, and operate globally.

F. Rapid technological change

Rules designed for conventional algorithms may become outdated when autonomous AI agents become commercially widespread.

24. Future Direction

The emerging model is likely to move from:

Reactive antitrust

toward:

Continuous competition governance.

This means authorities increasingly need to understand the architecture of markets, not simply individual transactions.

The OECD's recent work similarly emphasizes that AI can affect market structure, contestability, data access, entry barriers and competitive dynamics, making enforcement, advocacy and monitoring increasingly important.

25. Conclusion

Strategic competition oversight in self-organizing markets represents a shift from traditional supervision of human business decisions toward supervision of technological systems that organize economic behaviour.

The fundamental competition-law principles remain relevant:

• agreements should not become lawful merely because software implements them;
• dominance cannot automatically justify exclusionary conduct;
• platforms cannot necessarily use control over infrastructure to disadvantage rivals;
• algorithmic systems cannot be treated as a legal shield against competition liability.

The cases of Eturas, Topkins, Trod/GB eye, HRS, Booking.com and Google Shopping demonstrate different stages of this development.

The central future challenge is more difficult: how competition law should respond when autonomous algorithms independently learn, adapt and coordinate market behaviour without a traditional human agreement. Current international policy work treats this as an emerging area rather than a settled legal question.

Accordingly, effective competition policy for self-organizing markets should combine traditional antitrust enforcement, algorithmic auditing, data and infrastructure oversight, interoperability, continuous market monitoring, and carefully targeted ex ante regulation.