Algorithmic Pricing Risks in Competition Law

Detailed Explanation with at least 6 Case Laws

1. Introduction

Algorithmic pricing risks arise when businesses use computer algorithms, artificial intelligence (AI), machine learning, automated repricing systems, or pricing software to determine the prices of goods and services. These technologies can analyse demand, competitors’ prices, customer behaviour, inventory, and market conditions to adjust prices rapidly.

Algorithmic pricing can promote competition by reducing costs, improving efficiency, and allowing businesses to respond to changing market conditions. However, it can also create significant competition-law risks where algorithms facilitate price-fixing, tacit collusion, information exchange, discriminatory pricing, predatory pricing, or the exploitation of market power.

The central legal question is whether the use of an algorithm merely reflects independent commercial decision-making or instead facilitates conduct that restricts competition.

Key principle: Using an algorithm to set prices is not inherently unlawful. Liability generally depends on the applicable competition law, the conduct of the businesses involved, the evidence of coordination or exclusionary effects, and the relevant legal test.

2. Meaning and operation of algorithmic pricing

Algorithmic pricing refers to the automated use of computational rules or models to establish, recommend, or modify prices.

Common forms include:

Rule-based pricing: Prices change according to predefined rules, such as matching a competitor’s price.

Dynamic pricing: Prices change with demand, supply, capacity, time, or market conditions.

AI-based pricing: Machine-learning systems predict consumer demand, competitor reactions, or willingness to pay.

Algorithmic repricing: Software continuously monitors competitors and updates prices automatically.

Platform-mediated pricing: Online marketplaces or intermediaries influence prices through shared pricing tools, recommendations, or contractual restrictions.

Personalised pricing: Prices vary between customers based on data or predicted willingness to pay.

These systems can operate independently, use a common software provider, rely on shared market information, or be designed to implement an agreement between competitors. The legal risks differ substantially between these arrangements.

3. Principal algorithmic pricing risks

A. Algorithmic collusion and price-fixing

Competitors may use algorithms to implement an express agreement to fix prices or to coordinate conduct that would otherwise be determined independently. Automated systems can make an existing cartel more effective by monitoring compliance and detecting deviations quickly.

For example, competing suppliers might agree to maintain minimum prices and use automated software to enforce those prices across thousands of products.

Where an agreement or concerted practice is established, the use of software does not provide immunity from competition law.

B. Tacit coordination and parallel pricing

Algorithms can rapidly react to competitors’ prices, potentially producing persistent parallel pricing without direct communication.

This creates an important evidentiary distinction:

Parallel prices alone do not necessarily establish unlawful collusion.

Independent responses to publicly observable market conditions may be lawful.

Evidence of communications, information exchange, coordinated commitments, or other conduct inconsistent with independent competition may support an infringement finding.

The difficulty is particularly acute where algorithms learn that maintaining higher prices is more profitable than aggressive competition.

C. Hub-and-spoke coordination

A common pricing-software provider or digital platform may act as a central hub connecting competing businesses.

For example, several hotels may use the same pricing platform, which collects commercially sensitive information and recommends prices to all participating hotels. This arrangement creates risks if the information-sharing and software design facilitate coordination between competitors.

The use of the same provider is not, by itself, proof of an infringement. The authority must examine the nature of the information, the parties’ knowledge, their conduct, and the applicable legal requirements for establishing coordination.

D. Exchange of commercially sensitive information

Pricing algorithms may depend on information about competitors’ future prices, discounts, output, capacity, customer demand, or business strategies.

Exchanging such information directly, through a platform, or through a common service provider can reduce strategic uncertainty and weaken competition. Risk is particularly high where information is current, identifiable, detailed, and commercially sensitive.

E. Predatory pricing and exclusion of competitors

An algorithm may identify opportunities to price below cost to attract customers, weaken rivals, or deter new entry. Automated systems can execute such strategies across products and markets at a speed and scale that traditional monitoring may not capture.

Low prices are normally beneficial to consumers. A competition-law concern arises where the applicable legal test for predation or exclusionary conduct is satisfied, taking account of costs, market power, strategy, likely effects, and, where required, the prospect of recouping losses.

F. Personalised pricing and exploitation

AI systems may estimate customers’ willingness to pay using browsing history, purchasing patterns, location, device characteristics, or other data.

Personalised pricing is not automatically prohibited under competition law. It may nevertheless raise concerns where a dominant undertaking exploits market power, discriminatory conditions distort competition, or pricing practices violate consumer-protection or data-protection rules.

G. Price discrimination and market segmentation

Algorithms can distinguish between customer groups, geographical areas, or transaction types. Such differentiation can be commercially legitimate, for example where costs or demand differ. It may become problematic where it excludes rivals, exploits dependent trading partners, or breaches specific statutory prohibitions.

H. Algorithmic errors and feedback loops

Poorly designed models can produce artificial price spikes, unstable price movements, or mutually reinforcing price increases. Even where no deliberate coordination exists, these outcomes may trigger regulatory scrutiny, consumer-protection concerns, contractual disputes, and reputational damage.

The legal assessment must distinguish harmful outcomes from legally prohibited conduct: an unusual price movement does not, by itself, establish an antitrust violation.

4. Legal framework governing algorithmic pricing

A. United Kingdom

The principal legislation is the Competition Act 1998:

Chapter I prohibition: Restricts agreements and concerted practices that prevent, restrict, or distort competition.

Chapter II prohibition: Prohibits abuse of a dominant position.

Enterprise Act 2002: Provides additional enforcement mechanisms, including director-disqualification provisions and criminal cartel offences in applicable circumstances.

The Competition and Markets Authority (CMA) has made clear that automated repricing systems do not exempt businesses from ordinary competition rules. The central concern is whether the technology implements an unlawful agreement or facilitates coordination through competitively sensitive information.

B. European Union

Article 101 of the Treaty on the Functioning of the European Union (TFEU) prohibits qualifying anticompetitive agreements and concerted practices. Article 102 prohibits abuse of a dominant position.

Algorithmic pricing may be relevant to both provisions, depending on the conduct involved. A common software provider does not automatically establish an agreement between users; the authority must establish the legally required elements of coordination.

C. India

The Competition Act, 2002, is particularly relevant to algorithmic pricing in online marketplaces, ride-hailing platforms, and digital intermediaries.

Section 3: Prohibits specified anticompetitive agreements, including price-fixing arrangements.

Section 4: Addresses abuse of dominant position.

Section 19: Provides for inquiry into alleged contraventions and identifies relevant factors for assessing anticompetitive agreements and dominance.

Section 26: Governs the procedure for investigation and related orders at the relevant stage.

In India, the fact that a platform determines fares algorithmically does not automatically establish a cartel. The existence of an agreement or arrangement and the applicable statutory requirements remain central.

D. United States

Section 1 of the Sherman Act prohibits contracts, combinations, and conspiracies in restraint of trade, while Section 2 addresses monopolisation and attempted monopolisation.

Competitors cannot avoid liability for an unlawful price-fixing agreement merely by delegating its implementation to software. However, independent algorithmic pricing or parallel conduct alone does not necessarily establish the agreement required under Section 1.

5. Important case laws on algorithmic pricing risks

The following cases illustrate the legal principles applicable to algorithmic pricing. Some directly concern automated pricing; others establish broader rules on collusion, information exchange, parallel conduct, and exclusionary pricing.

Case 1: United States v. Topkins (2015)

Jurisdiction: United States.

Facts: Participants in online sales of posters used pricing algorithms after discussing and agreeing on how prices should be coordinated.

Legal issue: Whether software-assisted pricing could implement an unlawful price-fixing conspiracy.

Principle: The use of algorithms does not shield participants from antitrust liability when the technology implements an agreement to coordinate prices.

Significance: This is an important example of explicit algorithmic collusion. The technology made the agreement easier to implement and monitor, but the underlying infringement was the agreement itself.

Case 2: Trod Ltd. and GB eye Ltd. — Online Sales of Posters and Frames (2016)

Jurisdiction: United Kingdom.

Facts: Two competing online sellers agreed not to undercut one another on specified products sold through Amazon's UK website. They configured automated repricing software to implement the agreement.

Legal issue: Whether automated repricing used to enforce an agreement not to undercut competitors infringed competition law.

Principle: Automated pricing software cannot legitimise a price-fixing agreement that would be unlawful if implemented manually.

Outcome: The CMA found an infringement of the Chapter I prohibition and imposed a fine of £163,371 on Trod, following a settlement and cooperation discount. GB eye received leniency subject to the relevant conditions.

Significance: This is a leading enforcement example demonstrating that businesses remain responsible for the competitive effects of how they configure and use pricing software.

Case 3: Eturas UAB and Others v. Lietuvos Respublikos konkurencijos taryba (Case C-74/14, 2016)

Jurisdiction: European Union.

Facts: The dispute concerned an online travel-booking system whose administrator sent a system message concerning restrictions on discounts offered by participating travel agencies.

Legal issue: Whether agencies using the common platform could be treated as participating in a concerted practice involving a restriction on discounting.

Principle: Participation in a common digital system and awareness of a communication may be relevant to establishing concertation, but liability cannot automatically be inferred merely because an undertaking used the system. The applicable evidentiary requirements and opportunity to rebut the inference are important.

Significance: The case is particularly relevant to platforms where a software administrator can influence pricing behaviour across multiple independent businesses. It demonstrates why authorities must examine knowledge, participation, and actual market conduct rather than relying on common software use alone.

Case 4: Samir Agrawal v. Competition Commission of India (2021) 3 SCC 136

Jurisdiction: India.

Facts: The case concerned allegations that ride-hailing platforms Ola and Uber used algorithmically determined fares and thereby facilitated price coordination among drivers.

Legal issue: Whether the platform's role in determining fares established an anticompetitive agreement or hub-and-spoke arrangement among drivers.

Principle: Algorithmic fare determination and the existence of a common platform do not, without sufficient evidence of the legally required arrangement, establish a cartel.

Outcome: The Supreme Court of India upheld the rejection of the complaint, finding insufficient grounds to establish the alleged anticompetitive arrangement on the material before the authorities.

Significance: The judgment is important in distinguishing platform-mediated pricing from unlawful coordination. It also demonstrates that a common intermediary does not necessarily eliminate the need to establish an agreement between the relevant participants.

Case 5: Dole Food Company, Inc. v. European Commission (Case C-286/13 P, 2015)

Jurisdiction: European Union.

Facts: The case concerned coordination and exchanges of commercially sensitive information in the banana market.

Legal issue: Whether communications concerning future pricing intentions and market conduct could establish unlawful coordination.

Principle: Information exchanges that reduce uncertainty about competitors' future market behaviour may form part of an anticompetitive concerted practice.

Significance: Although not an AI-pricing case, the judgment is relevant where pricing algorithms rely on or transmit competitors' confidential pricing intentions. It helps explain why data-sharing arrangements between algorithm users may attract scrutiny.

Case 6: Anic Partecipazioni SpA v. Commission (Case C-49/92 P, 1999)

Jurisdiction: European Union.

Facts: The case concerned participation in a cartel and the evidentiary relationship between concertation and subsequent market conduct.

Legal issue: How the existence of a concerted practice and the connection between coordination and market behaviour should be assessed.

Principle: EU competition law recognises concerted practices as a distinct basis of liability. Where the relevant conditions are met, a rebuttable presumption may connect established concertation with subsequent market conduct.

Significance: This principle may be relevant when firms exchange pricing information through software and then alter their prices in a manner consistent with that exchange. The authority must still establish the required elements of the infringement.

Case 7: United States v. Apple Inc. (2013)

Jurisdiction: United States.

Facts: The case concerned arrangements between Apple and publishers relating to the pricing of e-books, including the use of agency agreements and most-favoured-nation provisions.

Legal issue: Whether arrangements that changed the distribution and pricing model in a market formed part of an unlawful conspiracy to raise prices.

Principle: Competition law examines the substance and context of commercial arrangements rather than treating a new pricing model as inherently lawful or unlawful.

Significance: The case is relevant to digital markets in which a platform sets contractual pricing rules, coordinates multiple suppliers, or changes incentives across a market. It is not a direct AI-pricing decision, but it illustrates the risks of platform-mediated price coordination.

Case 8: United States v. Socony-Vacuum Oil Co. (1940)

Jurisdiction: United States.

Facts: The case concerned a scheme involving purchases of surplus gasoline intended to influence market prices.

Legal issue: Whether a price-stabilisation arrangement among competitors could violate antitrust law.

Principle: Under US antitrust law, agreements among competitors to fix prices are generally unlawful regardless of whether the participants believe the arrangements stabilise the market or serve a commercially desirable purpose.

Significance: The principle remains relevant when algorithms are deliberately configured to maintain prices, suppress discounts, or implement a price-fixing agreement. The use of sophisticated software does not change the legal character of the underlying arrangement.

6. Comparative analysis of the cases

The direct cases are especially important because they demonstrate the distinction between using algorithms to compete and using them to implement or facilitate coordination.

7. Hypothetical examples of algorithmic pricing risks

Example 1: Competing online retailers

Two online retailers agree not to undercut each other. Each configures its repricing software to maintain a specified minimum price.

Legal assessment: This closely resembles the conduct in the Trod and GB eye case. The software is implementing an agreement between competitors, creating a substantial risk of liability.

Example 2: Hotels using a shared pricing platform

Several competing hotels provide non-public future pricing information to a software provider. The provider uses the information to generate recommendations that discourage price competition.

Legal assessment: The arrangement requires careful examination of information flows, contractual arrangements, participants' knowledge, and the provider's role. If the system facilitates coordination, competition-law exposure may arise.

Example 3: Independent retailers matching public prices

A retailer uses AI to observe publicly advertised prices and independently adjust its own prices to remain competitive.

Legal assessment: This is not inherently unlawful. The analysis changes if the retailer exchanges confidential information with competitors, participates in a common coordination mechanism, or uses the system to implement an unlawful agreement.

Example 4: Dominant platform using personalised pricing

A dominant digital platform uses customer-level data to charge different prices to different consumers.

Legal assessment: Personalisation alone does not establish an infringement. Relevant questions include whether the conduct exploits market power, excludes competitors, imposes prohibited discriminatory conditions, or breaches separate consumer or data-protection rules.

8. Enforcement challenges

Regulators face several difficulties when investigating algorithmic pricing.

Attribution: Determining whether the algorithm acted independently, followed human instructions, or implemented coordination.

Evidence: Recovering model logs, software configurations, communications, training data, and pricing histories.

Causation: Establishing whether coordination caused the observed pricing pattern.

Market definition: Assessing competition in multi-sided platforms and markets where services are supplied at zero monetary prices.

Intent and knowledge: Determining what businesses and software providers knew about the information and rules used by their systems.

Remedies: Designing interventions that restore competition without unnecessarily prohibiting beneficial automated pricing.

The challenge is not simply to identify unusual price movements. It is to establish the relevant legal infringement through evidence suited to the technology involved.

9. Compliance and risk-management framework

Businesses using pricing algorithms should adopt a structured compliance programme.

Independent pricing decisions: Ensure that prices are determined independently of competitors' confidential information.

Software-provider due diligence: Assess whether the provider pools, discloses, or uses competitors' commercially sensitive data.

Data governance: Restrict access to non-public pricing information and document permitted data flows.

Algorithmic audits: Test whether system rules encourage price matching, discount suppression, or coordinated responses.

Human oversight: Require appropriate review of high-risk pricing recommendations and significant changes to pricing logic.

Documentation: Maintain records of model objectives, pricing decisions, data sources, configurations, and compliance reviews.

Competition-law training: Train commercial teams, data scientists, and software providers to identify cartel risks.

Incident response: Establish a procedure for investigating suspicious price convergence, unexplained price increases, or inappropriate information exchange.

These controls should be proportionate to the firm's market position, the sensitivity of the information processed, and the degree of interaction with competitors.

10. Conclusion

Algorithmic pricing presents both opportunities and risks for competition. It can improve market responsiveness and reduce operating costs, but it can also make price-fixing easier to implement, information exchange more efficient, and exclusionary strategies more difficult to detect.

The case law establishes three essential propositions:

Technology does not excuse unlawful conduct. A cartel remains unlawful when implemented through automated repricing software.

Parallel prices are not sufficient by themselves. Authorities must establish the applicable legal elements of an agreement, concerted practice, or other infringement.

Digital intermediaries require careful scrutiny. Common pricing systems, data-sharing arrangements, and platform rules can facilitate coordination, but their legal significance depends on the evidence and circumstances.

Ultimately, effective competition-law enforcement should focus on the conduct, information flows, incentives, and market effects surrounding an algorithm—not merely on the fact that a business uses AI or automated pricing.