Competition Law and Public Algorithm Governance and Antitrust

1. Introduction

Public algorithm governance refers to the design, deployment, supervision and regulation of algorithmic systems used by governments, regulators, public authorities and public-sector entities.

Algorithms may be used for:

public procurement;

allocation of public resources;

taxation and compliance;

market surveillance;

competition-law enforcement;

price monitoring;

public-service allocation;

transport regulation;

licensing;

fraud detection;

ranking and eligibility decisions;

regulatory reporting.

From a competition-law perspective, algorithmic governance creates a dual problem.

First, private firms may use algorithms to restrict competition.

Second, public authorities may themselves design or operate algorithmic systems that affect competitive conditions.

The second category is particularly important because an algorithm used by a public authority can indirectly determine:

who obtains market access;

which suppliers qualify for contracts;

how bids are evaluated;

which businesses are investigated;

which firms receive regulatory approval;

how scarce public resources are allocated.

Therefore, competition law increasingly has to consider the relationship between algorithmic decision-making, administrative power, market structure and antitrust enforcement.

2. Meaning of Public Algorithm Governance

Public algorithm governance involves establishing rules concerning the entire lifecycle of an algorithm:

Design → Data → Training → Deployment → Decision → Monitoring → Audit → Review

A public authority might use an algorithm to rank suppliers in a government procurement process.

For example:

Supplier A + Supplier B + Supplier C

↓

Public procurement algorithm

↓

Risk score / quality score / price score

↓

Award decision

If the algorithm systematically disadvantages new entrants or favours incumbent suppliers, competition concerns may arise.

3. Why Competition Law Is Relevant

Algorithms can affect competition in at least five ways.

1. Market access

An algorithm can determine which businesses receive access to a public market.

2. Competitive neutrality

A public algorithm can favour state-owned or incumbent enterprises.

3. Information exchange

Algorithms can facilitate communication of commercially sensitive information.

4. Coordination

Competitors may use common algorithms to align prices or commercial behaviour.

5. Enforcement

Competition authorities increasingly use algorithms themselves to detect:

cartels;

bid rigging;

suspicious pricing;

mergers;

exclusionary conduct.

Thus, algorithms can simultaneously be objects of competition-law scrutiny and tools of competition-law enforcement.

4. Public Algorithms and Administrative Market Power

A public authority may not be a conventional commercial undertaking, but its decisions can substantially affect competition.

For example:

Government algorithm → supplier ranking → procurement allocation → market share

The algorithm may therefore become a regulatory gateway.

Potential problems include:

discriminatory scoring;

opaque eligibility rules;

exclusionary technical standards;

preferential treatment;

algorithmic barriers to entry;

automated regulatory decisions that unintentionally reinforce concentration.

Competition law may need to be considered alongside:

administrative law;

procurement law;

constitutional principles;

data-protection law;

transparency requirements.

5. Algorithms and the Relevant Market

Algorithmic systems can make traditional market definition more difficult.

A government may regulate a market using data generated by:

online platforms;

payment systems;

logistics networks;

digital marketplaces.

The algorithm may therefore operate across several interconnected markets.

For example:

Data market → algorithmic analytics → procurement market → downstream consumer market

Competition authorities should consider whether the algorithm creates competitive effects beyond the immediate activity for which it was designed.

6. Case Law: Eturas v Lietuvos Respublikos konkurencijos taryba

Court: Court of Justice of the European Union

Eturas is one of the most important cases concerning technology-mediated coordination.

A common online booking system was used by travel agencies. A system message introduced a restriction on discounts that could be offered.

The case raised the question of whether conduct implemented through a common digital system could constitute a concerted practice.

Principle

The use of software does not prevent competition law from applying.

A technological platform can become the mechanism through which competitors receive information or implement coordinated restrictions.

Relevance to public algorithm governance

A public procurement or regulatory algorithm could similarly influence multiple market participants.

The critical questions include:

who designed the algorithm;

what information it receives;

what information it communicates;

how participants respond;

whether the system creates discriminatory or coordinated outcomes.

7. T-Mobile Netherlands v Netherlands Competition Authority

Court: Court of Justice of the European Union

The case concerned information exchanged among competitors during a meeting.

The Court emphasised the importance of preserving independent competitive decision-making.

Algorithmic significance

Modern competitors do not necessarily exchange information through traditional meetings.

Information can be transmitted through:

software;

shared databases;

algorithms;

platforms;

automated pricing systems.

The underlying competition principle remains relevant:

Competitors should not replace independent decision-making with coordinated decision-making facilitated by information systems.

This is particularly important where public authorities operate platforms through which competitors submit commercially sensitive information.

8. AC-Treuhand v Commission

Court: Court of Justice of the European Union

The case concerned an undertaking that facilitated cartel activity without necessarily being a conventional producer of the cartelised product.

The Court's jurisprudence demonstrates that an undertaking can potentially attract competition-law liability where it knowingly contributes to anti-competitive coordination.

Algorithmic governance relevance

An algorithmic intermediary could potentially become important where it:

collects competitors' sensitive information;

communicates strategic information;

implements coordinated rules;

facilitates price coordination.

The technological character of the intermediary does not itself determine the competition-law analysis.

9. Google Shopping

European Commission / General Court

The Google Shopping litigation concerned Google's treatment of its comparison-shopping service within its general search results.

The case is significant for the competition-law discussion surrounding self-preferencing and algorithmic ranking.

Algorithmic significance

Search and recommendation algorithms can determine which competitors receive visibility.

An algorithm can therefore become a competitive gatekeeper.

Potential concerns include:

preferential ranking of affiliated services;

demotion of rivals;

manipulation of search visibility;

discriminatory access to users.

Public governance relevance

The same principle can arise where a government-controlled digital marketplace or procurement platform uses an algorithm to rank public suppliers.

The question becomes whether apparently neutral algorithmic criteria systematically favour particular participants.

10. Microsoft v Commission

General Court of the European Union

Microsoft concerned, among other matters, interoperability and access to information necessary for competing products to function effectively with Microsoft's dominant operating system.

Algorithmic governance relevance

Modern government algorithms frequently depend upon:

APIs;

databases;

authentication systems;

interoperability;

data exchanges.

If a dominant technology provider controls an infrastructure upon which a public algorithm depends, restrictions on interoperability may affect the competitive opportunities of other providers.

The case therefore illustrates the importance of technical access as a competition issue.

11. United Brands v Commission

Court of Justice of the European Union

United Brands remains a foundational authority concerning dominance.

The Court considered the ability of an undertaking with substantial economic strength to behave independently of competitive constraints.

Relevance to algorithmic markets

An algorithmic system can reinforce dominance through:

network effects;

data advantages;

economies of scale;

switching costs;

technological lock-in.

Therefore, when evaluating an algorithmically intensive market, authorities should not consider only traditional market shares.

12. Wouters v Algemene Raad van de Nederlandsche Orde van Advocaten

Court: Court of Justice of the European Union

The case concerned rules adopted by a professional regulatory body.

The Court recognised that a regulatory rule may restrict competition while nevertheless escaping prohibition where it is reasonably necessary for legitimate regulatory objectives within the relevant framework.

Public algorithm governance relevance

This principle is highly relevant where public authorities establish algorithmic rules affecting market participants.

An algorithmic regulation may restrict competitive freedom because it pursues objectives such as:

public safety;

consumer protection;

professional standards;

financial stability;

cybersecurity.

The competition analysis therefore needs to distinguish between:

unjustified competitive restrictions

and

restrictions reasonably connected with legitimate regulation.

13. Meca-Medina and Majcen v Commission

Court: Court of Justice of the European Union

The case concerned sporting rules and their relationship with competition law.

The Court emphasised that the existence of a regulatory objective does not automatically remove rules from competition law.

Instead, their restrictive effects and legitimate objectives must be assessed.

Algorithmic governance significance

The same reasoning can be relevant to algorithmic regulatory systems.

For example, a public algorithm may restrict market behaviour to achieve:

safety;

environmental objectives;

fraud prevention;

consumer protection.

The regulatory purpose does not automatically settle the competition question.

The design and proportionality of the restriction remain important.

14. MOTOE v Elliniko Dimosio

Court: Court of Justice of the European Union

The case concerned a body involved in organising motorcycling activities while also exercising regulatory powers over those activities.

The Court addressed the problem of a body possessing regulatory authority while participating in an economic activity.

Public algorithm governance relevance

This is particularly important where a public or quasi-public authority:

regulates a market;

operates infrastructure in that market; and

uses algorithms to make regulatory or commercial decisions.

A conflict can arise if the authority's regulatory system gives its own economic activity an advantage.

The principle is particularly relevant to:

state-owned digital platforms;

public procurement systems;

government-operated marketplaces;

public transport platforms.

15. CIF v Autorità Garante della Concorrenza e del Mercato

Court: Court of Justice of the European Union

The case concerned the interaction between state regulatory measures and competition law.

It illustrates that national regulatory structures can sometimes create or reinforce anti-competitive conditions.

Algorithmic significance

A public authority cannot necessarily avoid competition-law scrutiny simply because restrictive market conditions arise through regulation rather than through a private agreement.

For algorithmic governance, this raises the question:

Does the regulatory framework merely administer a competitive market, or does it structurally impose or facilitate restrictions on competition?

16. Public Procurement Algorithms

Public procurement is one of the most important applications of public algorithm governance.

An algorithm may evaluate:

price;

technical capability;

past performance;

financial stability;

risk;

sustainability;

cybersecurity.

This can increase efficiency and reduce administrative discretion.

However, competition risks may arise if the algorithm:

favours incumbent firms;

gives excessive weight to past contracts;

imposes unnecessary technical requirements;

excludes smaller suppliers;

creates opaque qualification thresholds.

17. Algorithmic Procurement and Bid Rigging

Algorithms can also facilitate bid-rigging detection.

A public authority can analyse:

identical bid patterns;

suspicious price sequences;

unusual bid rotation;

repeated winning patterns;

geographic allocation;

timing patterns.

For example:

Supplier A wins Tender 1
Supplier B wins Tender 2
Supplier C wins Tender 3
prices remain unusually similar.

An algorithm can identify this pattern for further investigation.

Importantly, an algorithmic flag is evidence for investigation, not automatically proof of a cartel.

18. Algorithmic Price Monitoring

Competition authorities can use algorithms to monitor markets for:

sudden price alignment;

parallel pricing;

unusual price movements;

capacity restrictions;

suspicious bidding.

However, parallel prices alone do not establish unlawful coordination.

Prices may move together because firms face:

common costs;

common demand;

regulation;

supply shocks;

identical market conditions.

This is consistent with the reasoning associated with the Wood Pulp litigation.

19. Algorithmic Collusion

Algorithms may potentially facilitate collusion in several ways.

Explicit coordination

Firms directly instruct algorithms to coordinate.

Facilitated coordination

A third-party algorithm provides information that facilitates coordination.

Signalling

Algorithms respond to publicly visible market information in ways that may make prices converge.

Autonomous coordination

More sophisticated systems may learn from market behaviour without direct human communication.

Competition law is still concerned with the underlying competitive conduct, rather than simply whether a human or machine performed the action.

20. Public Algorithms and Tacit Coordination

A government algorithm can unintentionally make coordination easier.

Suppose a regulator publishes highly detailed information about:

individual firms;

future prices;

capacity;

production;

inventories.

Competitors may then be able to predict each other's behaviour more accurately.

Consequently, transparency can sometimes have a dual effect:

greater regulatory transparency → better market information

but potentially:

excessive firm-specific information → reduced strategic uncertainty among competitors.

Public authorities therefore need to distinguish legitimate transparency from disclosure that could facilitate coordination.

21. Algorithmic Discrimination

A public algorithm may classify businesses differently.

For example:

Large incumbent supplier → low risk

New entrant → high risk

If the distinction is based on historical data rather than objective competitive criteria, the system may unintentionally reinforce incumbency.

Potential consequences include:

reduced entry;

higher compliance costs for entrants;

exclusion from procurement;

reduced innovation;

concentration.

This is particularly important because historical datasets often reflect the market structure that already exists.

22. Algorithmic Feedback Loops

One of the most significant problems is the algorithmic feedback loop.

Consider:

Existing market leader → receives more government contracts → generates more performance data → algorithm learns from data → incumbent receives higher score → incumbent receives more contracts.

This can produce self-reinforcing market concentration.

The problem does not necessarily require discriminatory intent.

A formally neutral algorithm can produce exclusionary effects if its training data and design systematically favour established firms.

23. Public Algorithms and Incumbency Bias

Public authorities may unintentionally encode incumbency into algorithms.

Examples include:

excessive weighting of previous government contracts;

reliance on historical default rates;

requiring large minimum turnover;

scoring established certifications more heavily;

treating absence of historical data as increased risk.

Such criteria may be administratively convenient but can make entry more difficult.

Competition-sensitive algorithm design should therefore consider contestability and entry.

24. State-Owned Enterprises

Algorithmic governance becomes particularly sensitive where a government simultaneously:

regulates a sector;

operates an algorithmic platform;

owns a commercial undertaking.

For example:

State regulator + state-owned logistics platform + government allocation algorithm.

Potential concerns include:

discriminatory access;

preferential data;

favourable ranking;

information advantages;

exclusion of private competitors.

The MOTOE jurisprudence illustrates why combining regulatory authority with economic activity can raise structural competition concerns.

25. Algorithms and Essential Facilities

An algorithmic infrastructure can potentially become economically important if market participants cannot reasonably operate without it.

Examples could include:

mandatory government procurement portals;

public identity systems;

regulated payment infrastructure;

public digital certification systems.

If access is legally or practically indispensable, discriminatory access may have significant competitive consequences.

However, the legal criteria for an abuse based on refusal of access remain demanding, as illustrated by Bronner and related jurisprudence.

26. Algorithmic Governance and Interoperability

Government algorithms frequently depend upon multiple databases.

For example:

Tax database ↔ Business registry ↔ Procurement database ↔ Competition database

If interoperability is restricted, certain firms may receive better access than competitors.

Competition concerns can therefore arise from:

API restrictions;

data silos;

discriminatory interfaces;

incompatible data formats;

preferential access.

The Microsoft jurisprudence provides a useful framework for understanding the competitive importance of interoperability.

27. Algorithmic Governance and Merger Control

Competition authorities can use algorithms to screen mergers.

Algorithms can identify:

concentration patterns;

overlapping businesses;

acquisition networks;

common ownership;

serial acquisitions;

potential killer acquisitions.

But automated screening should not replace substantive merger analysis.

An algorithmic flag should lead to human investigation, not automatic condemnation.

28. Algorithms in Competition-Law Enforcement

Competition authorities increasingly have potential uses for computational tools in:

Cartel detection

Identifying suspicious bidding patterns.

Market monitoring

Tracking prices and output.

Merger screening

Identifying concentration trends.

Document analysis

Processing large quantities of evidence.

Network analysis

Identifying relationships between firms.

Digital-market investigations

Mapping platform ecosystems.

These applications can substantially increase enforcement capacity.

29. Risks of Algorithmic Antitrust Enforcement

Public authorities must also govern their own algorithms carefully.

Potential risks include:

False positives

Lawful conduct is incorrectly flagged as suspicious.

False negatives

Sophisticated anti-competitive behaviour remains undetected.

Data bias

Historical enforcement data may reproduce past assumptions.

Lack of explainability

Businesses may not understand why they were selected for investigation.

Automation bias

Officials may place excessive reliance on algorithmic outputs.

Privacy concerns

Large-scale monitoring may involve sensitive commercial information.

30. Human Oversight

A strong public algorithm-governance framework should maintain:

Algorithmic assistance + human legal judgment

rather than:

Algorithmic decision + no meaningful review

Human oversight is particularly important when decisions affect:

market access;

licensing;

procurement;

investigations;

penalties;

merger review.

31. Transparency and Explainability

Transparency can operate at several levels.

System transparency

Who developed the algorithm?

Data transparency

What categories of data are used?

Methodological transparency

What factors influence the result?

Procedural transparency

How can an affected undertaking challenge the result?

Auditability

Can independent authorities test whether the system produces discriminatory effects?

Complete disclosure of source code is not always necessary.

Trade secrets, cybersecurity and intellectual-property considerations may justify limited disclosure.

32. Competition-Neutral Algorithm Design

Public authorities can incorporate competition considerations into algorithm design.

Criterion 1 — Neutrality

Criteria should not unnecessarily favour incumbents.

Criterion 2 — Contestability

New entrants should have a reasonable opportunity to qualify.

Criterion 3 — Interoperability

Systems should use open or accessible interfaces where appropriate.

Criterion 4 — Reviewability

Businesses should have mechanisms to challenge adverse decisions.

Criterion 5 — Auditability

The system should be periodically tested for systematic competitive effects.

Criterion 6 — Data minimisation

Only competitively relevant information should be collected and disclosed.

33. Algorithms and Public Procurement

A competition-sensitive procurement algorithm should avoid unnecessarily relying upon:

incumbent status;

historical market share;

previous government contracts alone.

Instead, evaluation can focus on objectively relevant factors such as:

price;

technical performance;

quality;

financial capacity;

delivery capability;

legally relevant compliance.

This can make procurement more contestable.

34. Algorithmic Standards and Competition

Public authorities may establish technical standards through algorithms or automated systems.

Standards can create efficiencies by:

improving compatibility;

reducing transaction costs;

improving safety.

But standards can also affect which technologies survive.

A standard that effectively excludes competing technologies may raise competition concerns, particularly where:

participation is restricted;

alternative technologies are excluded without justification;

dominant firms control the standard-setting process.

The Rambus and Huawei v ZTE jurisprudence illustrates the broader importance of standards and intellectual property.

35. Public Algorithm Governance and Section 3 of the Indian Competition Act

Under Section 3 of the Competition Act, 2002, agreements that cause or are likely to cause an appreciable adverse effect on competition may attract scrutiny.

Algorithmic systems could become relevant where competitors use a common system to:

exchange sensitive information;

coordinate prices;

coordinate bids;

allocate customers;

restrict output.

The fact that the coordination occurs electronically rather than through a traditional written agreement does not fundamentally change the competition-law analysis.

36. Public Algorithm Governance and Section 4

Section 4 concerns abuse of dominant position.

A dominant technology provider supplying an algorithmic infrastructure could potentially be examined where it:

denies access;

discriminates between users;

ties services;

leverages dominance;

restricts market access.

The existence of an algorithm alone does not establish dominance or abuse. The relevant market, dominance and competitive effects must be established separately.

37. Public Regulatory Measures and Competition

Public authorities can influence competition through:

licensing;

technical standards;

procurement requirements;

certification;

data-access rules;

algorithmic scoring.

Competition law may therefore need to operate alongside regulatory law.

The cases involving Wouters, MOTOE and CIF demonstrate the broader principle that the relationship between public regulation and competition law can be complex.

38. Key Case-Law Table

39. Core Principles Emerging From the Case Law

Several principles emerge.

1. Technology does not remove competition-law responsibility

A restriction implemented through software can have the same competition significance as one implemented through a conventional contract.

2. Digital infrastructure can become a competitive bottleneck

Control over interoperability, data or technical interfaces can affect downstream competition.

3. Information matters

Algorithms can dramatically reduce the uncertainty that competitors face about each other's behaviour.

4. Regulation does not automatically eliminate competition concerns

A legitimate regulatory objective can coexist with restrictive competitive effects.

5. Regulatory power and economic activity require particular care

Where a public authority also participates economically, institutional conflicts can arise.

6. Algorithmic neutrality requires more than neutral wording

A formally neutral algorithm can produce systematically exclusionary outcomes through its data and design.

40. Recommended Public Algorithm Governance Framework

A competition-sensitive framework can be organised around eight principles:

1. Competition impact assessment
Assess potential effects on market structure before deployment.

2. Data governance
Identify what commercial information is collected and disclosed.

3. Non-discrimination
Test whether equivalent firms receive materially different treatment.

4. Interoperability
Avoid unnecessary technical barriers.

5. Contestability
Evaluate whether new entrants can realistically participate.

6. Explainability
Provide sufficient reasons for affected businesses to understand decisions.

7. Independent audit
Periodically test the system for systematic competitive effects.

8. Human review
Maintain meaningful human and legal review of consequential decisions.

41. Conclusion

Public algorithm governance is increasingly becoming a competition-law issue because algorithms can shape the structure of markets rather than merely automate administrative tasks.

The major antitrust concerns include:

algorithmic discrimination;

incumbency bias;

barriers to entry;

algorithmically facilitated collusion;

strategic information disclosure;

self-preferencing;

interoperability restrictions;

data-driven market power;

preferential treatment of state-owned enterprises;

algorithmic procurement exclusion;

automated merger and cartel screening; and

lack of transparency and meaningful review.

The cases of Eturas, T-Mobile Netherlands, AC-Treuhand, Google Shopping, Microsoft, Bronner, IMS Health, Wouters, Meca-Medina, MOTOE and CIF collectively demonstrate that competition law must examine both the technological mechanism and the underlying economic and regulatory conduct.

The central principle is that public algorithms should be designed not merely for administrative efficiency, but also with attention to competitive neutrality, market access, contestability, interoperability, transparency and reviewability.