Competition Law and Algorithmic Decision Infrastructure Competition

1. Introduction

“Algorithmic decision infrastructure” refers to the digital systems that businesses increasingly rely on to make commercial decisions. These systems can include pricing algorithms, recommendation engines, ranking systems, bidding software, demand-forecasting tools, revenue-management systems, automated procurement systems, matching systems, and AI-based decision platforms.

From a competition-law perspective, the important issue is not simply that a company uses an algorithm. Algorithms can improve efficiency, reduce costs, improve forecasting, and make markets more responsive. Competition concerns arise when control over the infrastructure changes the competitive process—for example, when rivals use the same decision-making intermediary, when commercially sensitive information is pooled, when an infrastructure provider controls access to an important market, or when algorithms are designed or used to coordinate competitive behaviour.

Existing competition law generally addresses these problems through rules concerning:

• anticompetitive agreements and concerted practices;
• information exchange;
• hub-and-spoke coordination;
• abuse of dominance or monopolisation;
• discriminatory access and self-preferencing;
• exclusionary conduct;
• interoperability and switching restrictions; and
• mergers involving important data, software, or technological infrastructure.

The key principle is that automation does not normally place commercial conduct outside competition law. The legal analysis instead focuses on who controls the system, what information enters it, how firms interact through it, and what effects the arrangement has on competition.

2. What Is Algorithmic Decision Infrastructure?

Algorithmic decision infrastructure can be understood as the technological layer through which commercial decisions are generated, communicated, recommended, or implemented.

Consider a market containing 100 independent sellers. Traditionally, each seller may determine its own prices using its own employees, information, and strategy.

Suppose instead that 70 sellers subscribe to the same AI-based revenue-management platform. Each seller supplies information concerning sales, prices, inventories, customer demand, and capacity. The platform processes those inputs and sends recommended prices back to participating businesses.

The platform has therefore become part of the market's decision infrastructure.

That situation is not automatically unlawful. However, competition authorities may investigate whether the infrastructure reduces genuine independent decision-making.

The basic competitive expectation is:

Firm A → independent commercial decision

Firm B → independent commercial decision

Firm C → independent commercial decision

A potentially problematic structure may instead become:

Firm A → common algorithm

Firm B → common algorithm

Firm C → common algorithm

Common algorithm → commercially important recommendations to A, B and C

The legal question becomes whether the common infrastructure facilitates coordination, exclusion, discrimination, or market power.

3. Independent Decision-Making as a Competition Principle

Competition law normally expects competitors to determine their market behaviour independently.

This does not prevent businesses from intelligently observing market conditions. A retailer may observe publicly available prices and respond competitively.

The concern becomes stronger where businesses exchange non-public strategic information concerning matters such as:

• future prices;
• planned discounts;
• production capacity;
• inventories;
• customer allocation;
• future output;
• bids;
• margins;
• planned commercial strategy.

An algorithm does not necessarily change the legal character of the information.

For example, competitors should not normally be able to escape competition rules merely by replacing a direct exchange:

Competitor A → sensitive information → Competitor B

with an indirect structure:

Competitor A → algorithm/platform → Competitor B.

This principle is particularly important for common pricing and revenue-management systems.

4. Main Competition-Law Concerns

A. Algorithmic Price Coordination

One of the clearest risks involves competitors deliberately using algorithms to implement an agreed pricing strategy.

Competitors might agree upon:

• minimum prices;
• maximum discounts;
• pricing formulas;
• margins;
• price increases; or
• rules governing automated repricing.

If the underlying agreement constitutes price fixing, using software to execute the agreement ordinarily does not make the arrangement lawful.

A classic modern example is United States v. David Topkins.

The U.S. Department of Justice brought a criminal horizontal price-fixing case involving posters sold through an online marketplace. According to the case materials, competitors agreed to coordinate prices and used algorithmic pricing software as part of implementing their agreement. Topkins entered a guilty plea.

The significance is straightforward:

Human cartel + algorithmic implementation remains a cartel problem.

5. Common Algorithms as Coordination Hubs

A more complicated problem arises when competitors do not communicate directly but instead rely upon the same intermediary.

Imagine:

Retailer A → Platform X

Retailer B → Platform X

Retailer C → Platform X

and Platform X collects commercially sensitive information and produces recommendations affecting all three retailers.

This resembles the traditional competition-law concept of a hub-and-spoke arrangement.

The platform or software provider can function as the hub, while competing users operate as the spokes.

Whether this amounts to an unlawful agreement depends on the applicable jurisdiction and evidence concerning knowledge, communication, participation, and common understanding.

Traditional hub-and-spoke jurisprudence is therefore highly relevant to modern algorithmic markets.

6. Case Law 1 — United States v. David Topkins

United States v. David Topkins, U.S. District Court, Northern District of California (2015).

This is one of the most directly relevant algorithmic competition cases.

Topkins was involved in selling posters through an online marketplace. U.S. authorities alleged an agreement among competing sellers to fix prices. Algorithmic pricing software was used to coordinate aspects of the pricing arrangement.

Topkins pleaded guilty to horizontal price fixing.

Competition-law significance

The case demonstrates that an algorithm can operate as the implementation mechanism of a traditional cartel.

Competition law therefore focuses on the economic arrangement rather than whether prices were entered manually.

A company cannot normally defend explicit collusion simply by saying:

“The computer determined the final price.”

Where the competitors agreed on the mechanism producing those prices, conventional cartel principles may apply.

7. Case Law 2 — Eturas UAB and Others v Lithuanian Competition Council

Case C-74/14, Eturas UAB and Others v Lietuvos Respublikos konkurencijos taryba, Court of Justice of the European Union, 2016.

This is one of the most important European cases for understanding common digital infrastructure.

Travel agencies used the same E-TURAS online booking system.

The system administrator introduced an automatic restriction affecting discounts available through the system and communicated through the system concerning the restriction.

The dispute concerned whether participating travel agencies could be treated as participating in a concerted practice.

The Court emphasized important evidentiary protections. Mere dispatch of a system message was not, by itself, sufficient to establish that every agency participated. Knowledge and the possibility of rebutting presumptions remained important.

Importance for algorithmic infrastructure

Eturas demonstrates that a shared digital platform can become the mechanism through which competitive parameters are coordinated.

The case is especially important because the coordination mechanism was technological rather than a conventional cartel meeting.

Modern equivalents could potentially involve:

• pricing dashboards;
• common marketplace software;
• automated discount engines;
• revenue-management systems; or
• common AI decision platforms.

8. Case Law 3 — VM Remonts v Competition Council

Case C-542/14, VM Remonts and Others v Konkurences padome, Court of Justice of the European Union, 2016.

This case concerned the circumstances in which anticompetitive conduct carried out through an independent service provider can be attributed to a company.

The Court explained that the conduct of an independent service provider cannot automatically be attributed to its customer.

However, attribution can arise where, among other circumstances:

1. the provider effectively acts under the undertaking's direction or control;
2. the undertaking knows about the anticompetitive objectives and intends to contribute to them; or
3. it could reasonably foresee the anticompetitive conduct and was prepared to accept that risk.

Importance for algorithms

This principle can be highly significant where businesses outsource commercial decisions to:

• algorithm developers;
• consultants;
• pricing platforms;
• AI providers; or
• common data-processing intermediaries.

Using an independent technology provider does not automatically make every customer responsible for everything the provider does.

At the same time, outsourcing cannot necessarily provide a shield where the legally required knowledge, control, participation, or acceptance of risk is established.

9. Case Law 4 — Interstate Circuit v United States

Interstate Circuit, Inc. v. United States, 306 U.S. 208 (1939).

This case predates computers but provides an important foundation for analysing modern platform-based coordination.

A theatre operator communicated proposed restrictions to multiple film distributors. The Supreme Court upheld the finding of an unlawful arrangement based on the circumstances surrounding the distributors' participation and implementation of the restrictions.

The structure later became a major reference point in discussions of hub-and-spoke conspiracies.

Algorithmic relevance

Replace the historical intermediary with a modern software platform:

Platform = hub

Competing businesses = spokes

The fundamental competition question remains whether there is sufficient evidence of an agreement or concerted arrangement connecting the participants.

Therefore, traditional conspiracy doctrine can remain relevant even where coordination is organised through digital infrastructure.

10. Case Law 5 — United States v Apple Inc.

United States v. Apple Inc., 952 F. Supp. 2d 638 (S.D.N.Y. 2013), affirmed 791 F.3d 290 (2d Cir. 2015).

The case concerned the distribution and pricing of e-books.

The courts found Apple liable under Section 1 of the Sherman Act in connection with a conspiracy involving major publishers to raise e-book prices.

Although this was not an AI case, it is highly relevant to algorithmic infrastructure because it demonstrates how a central intermediary can connect competitors participating in coordinated market behaviour.

Algorithmic relevance

A modern intermediary does not have to be a traditional distributor.

It could potentially be:

• an online marketplace;
• cloud platform;
• pricing-software provider;
• procurement system;
• advertising exchange; or
• AI decision engine.

The legal issue remains whether evidence establishes the agreement required by competition law rather than merely parallel use of the same technology.

11. Case Law 6 — Google Shopping

Google LLC and Alphabet Inc. v European Commission, Case T-612/17, General Court, 2021, with subsequent appellate proceedings.

Google Shopping concerns a different aspect of algorithmic infrastructure: dominant control over ranking and access infrastructure.

The European Commission's case concerned Google's treatment of its own comparison-shopping service relative to competing comparison-shopping services in general-search results.

The General Court upheld the Commission's finding concerning Google's favourable positioning and display of its own comparison-shopping service and the corresponding treatment of competing comparison-shopping services.

Importance

Algorithmic competition concerns are therefore not limited to collusion.

A company controlling an important decision or ranking infrastructure may potentially influence:

• which businesses consumers see;
• product rankings;
• market access;
• traffic allocation;
• recommendations; and
• commercial visibility.

Where the infrastructure owner possesses the legally required degree of market power, competition law may examine whether control over that infrastructure is being used to exclude or disadvantage competitors.

12. Case Law / Enforcement Development 7 — RealPage Litigation

The U.S. litigation concerning RealPage represents an especially important contemporary development in algorithmic pricing.

In August 2024, the U.S. Department of Justice and participating state attorneys general sued RealPage.

The government alleged that competing landlords supplied non-public, competitively sensitive rental information to RealPage's revenue-management system and that its software generated recommendations concerning apartment pricing and related terms.

The government alleged violations of Sections 1 and 2 of the Sherman Act. These are allegations in ongoing litigation rather than a final judicial determination that all alleged conduct occurred or violated the law.

Importance

RealPage illustrates the central modern question:

Can a common algorithmic intermediary become infrastructure through which competitors' commercially sensitive information influences their decisions?

That question is likely to remain significant for competition law involving AI and shared pricing technologies.

13. Information Pooling Through Algorithms

Algorithms require data.

Consequently, competition authorities may examine not only the algorithm's output but also its inputs.

Suppose competing hotels provide a common platform with:

• current occupancy;
• future reservations;
• expected demand;
• planned promotions;
• confidential room prices; and
• future capacity.

The platform aggregates the information and generates recommended prices.

The legal analysis may ask:

What information is shared?

Public historical information creates different concerns from confidential forward-looking competitor information.

How detailed is it?

Company-specific information can present greater risks than sufficiently aggregated information.

How current is it?

Real-time information can potentially reduce strategic uncertainty more strongly than old historical information.

Who receives the results?

A genuinely aggregated market report can raise different issues from recommendations generated using individual competitors' confidential strategies.

14. Algorithmic Parallelism Is Not Automatically Collusion

An important distinction must be maintained.

Suppose two competing airlines independently purchase similar software.

Both systems observe publicly available market information and independently reach similar pricing decisions.

Similar prices alone do not necessarily prove an anticompetitive agreement.

Competition law traditionally distinguishes between:

Independent parallel conduct

and

coordinated conduct.

Algorithms complicate the evidentiary problem because sophisticated systems can react rapidly to one another.

Authorities and courts therefore may need to examine evidence such as:

• communications between competitors;
• communications with the algorithm provider;
• contractual arrangements;
• system architecture;
• data-sharing arrangements;
• algorithm design;
• knowledge of other participants;
• monitoring mechanisms; and
• departures from ordinary independent commercial behaviour.

15. Algorithmic Self-Preferencing

Another issue arises where the company controlling the algorithm also competes with businesses affected by its decisions.

Consider a marketplace that sells its own products while hosting independent sellers.

Its recommendation algorithm determines which products receive prominent placement.

Potential concerns arise if a dominant infrastructure operator systematically designs ranking rules that advantage its own downstream operations in circumstances satisfying the applicable legal test for abuse or monopolisation.

Google Shopping demonstrates why ranking infrastructure can become a competition issue.

The key question is therefore not simply:

“Is the algorithm biased?”

Competition law asks more specific questions concerning market power, conduct, competitive effects, objective justification, and the applicable statutory test.

16. Algorithmic Access Discrimination

Decision infrastructure can also operate as a market gateway.

Examples include:

• advertising exchanges;
• app-distribution systems;
• search engines;
• payment infrastructure;
• digital marketplaces;
• cloud ecosystems;
• procurement platforms.

If businesses depend heavily upon that infrastructure, discriminatory algorithmic access rules can potentially affect competition.

Authorities may investigate whether similarly situated businesses receive materially different treatment and whether the difference reflects legitimate technical or commercial considerations or exclusionary conduct by a dominant undertaking.

17. Data Advantages and Entry Barriers

Algorithmic infrastructure can generate strong feedback loops.

For example:

More users → more data

More data → better algorithm

Better algorithm → better service

Better service → more users

This process can be competitively beneficial.

However, under certain market conditions it can also make entry difficult.

A new competitor may require substantial:

• training data;
• computing resources;
• customer adoption;
• historical transaction information;
• integrations; and
• technical expertise.

Competition authorities therefore increasingly examine whether control over important datasets or technological infrastructure contributes to durable market power.

Possessing superior data is not itself generally unlawful. The competition issue usually concerns how market power was acquired, maintained, or exercised.

18. Interoperability and Switching

Infrastructure competition can also depend upon interoperability.

Suppose a business adopts a decision platform and accumulates years of commercial data inside it.

Changing providers becomes difficult if:

• data cannot be exported;
• APIs are restricted;
• technical interfaces are proprietary;
• contractual terms restrict migration;
• complementary products only work inside the original ecosystem.

These circumstances can increase switching costs.

High switching costs can strengthen an incumbent's position because customers may remain even when alternative technology becomes available.

Competition analysis may therefore examine whether restrictions are objectively necessary for security or technical integrity or instead function primarily as exclusionary barriers.

19. Multi-Market Algorithmic Infrastructure

A particularly significant issue arises where one decision provider serves numerous industries.

For example, one AI infrastructure company might provide:

• retail pricing;
• hotel revenue management;
• airline forecasting;
• logistics optimisation;
• advertising bidding; and
• procurement optimisation.

Its competitive significance may then extend beyond any single product market.

Authorities may need to analyse different layers:

Layer 1 — computing infrastructure

Layer 2 — data infrastructure

Layer 3 — algorithm/model infrastructure

Layer 4 — decision software

Layer 5 — downstream commercial markets

Market power at one layer can sometimes affect competition at another.

20. Merger Control

Algorithmic decision infrastructure can also become relevant when technology companies merge.

Authorities may investigate whether a transaction combines:

• competing algorithms;
• unique datasets;
• AI models;
• cloud infrastructure;
• important customer relationships;
• complementary software ecosystems.

The central question remains whether the merger may substantially reduce competition under the relevant jurisdiction's merger rules.

For example, authorities might examine whether the transaction removes an emerging algorithmic competitor or gives the merged firm control over infrastructure required by downstream competitors.

21. Responsibility Cannot Simply Be Delegated to AI

A significant governance principle follows from cases such as VM Remonts and Eturas.

Businesses should not assume that outsourcing commercial decisions to technology automatically removes competition-law responsibility.

Relevant questions can include:

• Who selected the algorithm?
• Who determined its objectives?
• What information was supplied?
• Did the company know competitors were participating?
• Could users modify recommendations?
• Were recommendations automatically implemented?
• What communications occurred between participants?
• Did the provider disclose how competitor information was being used?

These factual questions can determine whether competition-law liability exists.

22. Compliance Measures

Businesses using algorithmic decision infrastructure can reduce competition risks through appropriate compliance controls.

Useful measures include:

1. keeping competitively sensitive information separated from competitors;
2. reviewing third-party pricing and optimisation software before deployment;
3. understanding what competitor information enters shared models;
4. avoiding agreements concerning automated pricing parameters with competitors;
5. preserving genuine independent commercial decision-making;
6. auditing algorithms for potentially exclusionary rules where the business possesses significant market power;
7. documenting legitimate reasons for ranking and access criteria;
8. reviewing interoperability and data-portability restrictions; and
9. ensuring competition-law personnel understand how automated systems actually operate.

Simply telling employees not to communicate directly with competitors may be insufficient if the same sensitive information is indirectly exchanged through common infrastructure.

23. Summary of Major Authorities

24. Overall Legal Position

Competition law does not generally prohibit businesses from using algorithms, AI, shared software, or automated commercial tools.

The competition problem arises when algorithmic infrastructure becomes a mechanism for coordination, exclusion, discriminatory market access, misuse of competitively sensitive information, or maintenance of market power.

Three distinctions are particularly important.

First, algorithmic similarity is not necessarily algorithmic collusion. Competing algorithms can independently reach similar decisions.

Second, technology does not neutralise an underlying anticompetitive agreement. Where competitors intentionally coordinate prices and use software to execute that arrangement, traditional cartel law remains relevant.

Third, control over decision infrastructure can itself become competitively significant. A dominant platform controlling rankings, recommendations, data, interfaces, or market access may potentially affect downstream competition even without horizontal collusion.

The cases from Interstate Circuit and Apple to Eturas, VM Remonts, Topkins, Google Shopping, and the ongoing RealPage litigation show how traditional competition-law concepts are being applied to increasingly automated markets. The technology may be new, but the central legal objective remains familiar: preserving independent competitive decision-making and preventing firms from using agreements or market power to distort the competitive process.