1. Introduction

Competition Law and Theories of Algorithmic Governance examines how competition authorities, courts, governments, and market participants should regulate markets in which algorithms and artificial intelligence (AI) increasingly influence prices, output, consumer choice, market access, and business strategy.

Traditional competition law was developed largely around human decisions, written agreements, observable business conduct, and conventional market structures. Algorithmic markets introduce new challenges: competing firms can use automated pricing systems, platforms can rank their own products, and AI systems can facilitate coordination without requiring an explicit agreement between competitors.

Algorithmic governance refers to the legal, institutional, technical, and economic mechanisms used to guide, monitor, and control algorithmic decision-making. In competition law, its purpose is to ensure that algorithms improve efficiency and innovation without facilitating collusion, exclusionary conduct, discriminatory access, or the abuse of market power.

The central question is:

How can competition law preserve fair and effective market rivalry when important commercial decisions are increasingly made or influenced by algorithms?

The answer requires a combination of established antitrust principles, economic analysis, technological auditing, institutional accountability, and effective remedies.

2. Meaning and scope of algorithmic governance

Algorithmic governance involves setting rules and safeguards for the design, deployment, operation, and supervision of automated decision-making systems.

In competition law, it operates at three principal levels.

A. Legal governance

Competition statutes, merger control, prohibitions on cartels and abuse of dominance, procedural safeguards, and judicial review.

B. Economic governance

Market definition, assessment of market power, price effects, entry barriers, switching costs, network effects, and innovation incentives.

C. Technical governance

Algorithm audits, access controls, monitoring logs, data governance, testing, explainability, and safeguards against automated coordination.

These levels should operate together. Technical complexity does not remove a company's legal responsibility, while an unusual algorithmic outcome alone does not establish an infringement.

3. Major theories of algorithmic governance in competition law

3.1 The theory of algorithmic collusion

Algorithmic collusion theory examines whether automated systems facilitate coordination between competing businesses.

Algorithms may monitor rivals' prices, rapidly adjust offers, detect deviations from a common pricing pattern, and respond to competitors' conduct faster than human decision-makers.

Three situations should be distinguished:

Explicit algorithmic collusion: Competitors agree to coordinate prices or other competitive parameters and use algorithms to implement their agreement.

Hub-and-spoke coordination: A common platform, software provider, or intermediary facilitates coordination among businesses that compete with one another.

Tacit algorithmic coordination: Independently operating algorithms learn to follow similar pricing patterns without an explicit agreement.

The first two situations may fall within established prohibitions on agreements or concerted practices, depending on the evidence and applicable law. The third is more difficult: parallel pricing or machine learning alone does not necessarily prove an unlawful agreement.

Governance implication: Authorities should investigate communications, software procurement, shared pricing instructions, data exchanges, system design, and evidence of coordination rather than treating similar algorithmic outputs as conclusive proof of collusion.

3.2 The theory of algorithmic tacit coordination

This theory focuses on markets in which firms independently use algorithms that can quickly react to competitors.

Repeated interaction may allow algorithms to maintain elevated prices, punish discounting, or rapidly restore a common pricing pattern. Such outcomes are more plausible in some market conditions than others.

Relevant conditions include:

A small number of competitors.

Frequent and transparent price changes.

Homogeneous or easily comparable products.

High barriers to entry.

Predictable demand and rapid detection of deviations.

However, algorithmic pricing can also intensify competition, reduce costs, and produce lower prices. Therefore, a competition authority must distinguish harmful coordination from legitimate independent adaptation.

Governance implication: Market monitoring should identify suspicious patterns and economic effects, but enforcement should remain grounded in the applicable legal requirements for proving an infringement.

3.3 The theory of algorithmic abuse of dominance

A dominant undertaking may use algorithms to strengthen or exploit its market position.

Examples include:

Ranking its own products more favourably than competing products.

Using privileged platform data to disadvantage business users.

Restricting interoperability or access to essential platform functions.

Algorithmically excluding rivals from search, recommendation, or distribution channels.

Personalising offers or prices in ways that may exploit market power or foreclose competitors.

Not every instance of self-preferencing, differential ranking, or personalised pricing is unlawful. The legal assessment depends on market power, the nature of the conduct, competitive effects, objective justifications, and the relevant jurisdiction's rules.

Governance implication: Authorities should examine whether a system creates exclusionary effects, distorts access to customers, or protects market power by making effective competition unnecessarily difficult.

3.4 The theory of data-driven market power

Data can become a competitive advantage when it is difficult for rivals to replicate and is combined with scale, network effects, computing infrastructure, or superior predictive capabilities.

A company may gain a reinforcing advantage through the following cycle:

More users and transactions

More data and stronger feedback

Better predictions and product optimisation

Stronger market position and user retention

The cycle can reinforce itself when entry barriers and switching costs prevent effective rivals from emerging.

The legal question is not whether a business possesses a large dataset, but whether its data advantages contribute to durable market power or unlawful exclusion.

3.5 The theory of platform gatekeeping

Digital platforms may control access to customers, app distribution, search visibility, advertising, payment systems, or transaction infrastructure.

Algorithms determine how products are ranked, recommended, priced, and presented. A platform may consequently influence both its own competition with business users and competition among those users.

Governance tools may include ranking transparency, non-discrimination safeguards, interoperability, data portability, and restrictions on the misuse of commercially sensitive information.

Such interventions should be tailored to the identified competitive harm and balanced against privacy, cybersecurity, and legitimate product-design interests.

3.6 The theory of algorithmic accountability

Algorithmic accountability holds that businesses should remain answerable for commercially significant decisions made by their automated systems.

Relevant safeguards include:

Assigning identifiable responsibility for algorithm deployment.

Maintaining records of material system changes and pricing instructions.

Testing systems for coordination and exclusion risks.

Reviewing third-party software and data dependencies.

Preserving evidence for investigations.

Establishing procedures to suspend or modify problematic systems.

A company should not escape competition-law liability simply because a machine generated the relevant decision. At the same time, liability must be established under the applicable legal test rather than inferred solely from automation.

3.7 The theory of adaptive and risk-based regulation

Algorithmic markets evolve rapidly, making static regulatory approaches potentially inadequate.

Risk-based governance prioritises markets and systems where the potential for consumer harm or competitive distortion is greatest. Adaptive governance allows authorities to update monitoring methods as technology and market structure change.

For example, an authority could require enhanced reporting from a dominant digital platform where evidence indicates a significant risk of discriminatory ranking, while using ordinary enforcement mechanisms for lower-risk systems.

Adaptive governance must still respect legal certainty, proportionality, confidentiality, procedural fairness, and judicial oversight.

4. Important case laws on algorithmic governance and competition law

The following cases illustrate how established competition-law principles apply to algorithmic pricing, digital platforms, data-driven market power, and automated decision-making. Some directly concern algorithmic conduct; others provide foundational principles relevant to algorithmic governance.

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

Court: Court of Justice of the European Union.

Facts: The operator of a shared online travel-booking system communicated a restriction on the discounts that participating travel agencies could offer. Technical changes to the system made the discount restriction easier to implement.

Legal issue: Whether businesses using the common platform could be considered participants in an unlawful concerted practice.

Decision and principle: The Court explained that awareness of an anticompetitive initiative communicated through the platform could support an inference of participation, subject to the evidentiary requirements and the possibility of rebutting that inference. Mere receipt of a message could not automatically establish liability without the necessary assessment of awareness and other evidence.

Relevance to algorithmic governance: This is a particularly important case because it demonstrates how a shared digital system can facilitate coordination between independent businesses. Platform operators and users should assess whether system settings, communications, and automated restrictions implement or facilitate a common anticompetitive arrangement.

Governance lesson: Maintain auditable records of pricing instructions, system changes, communications, and decisions to participate in or reject potentially anticompetitive arrangements.

Case 2: United States v. Topkins (N.D. Cal., 2015)

Court: United States District Court for the Northern District of California; criminal antitrust proceedings.

Facts: An online seller participated in an agreement with competing sellers to fix prices for certain posters sold through an online marketplace. The sellers used automated pricing software to implement their agreement.

Legal issue: Whether the use of automated repricing software altered the application of established prohibitions on price fixing.

Outcome and principle: The matter resulted in a guilty plea. It demonstrated that competitors cannot avoid cartel liability by implementing an unlawful agreement through software rather than manually adjusting prices.

Relevance to algorithmic governance: Pricing algorithms may make collusive arrangements easier to execute, monitor, and maintain. However, the decisive legal element in this case was the underlying agreement, not simply the use of automated pricing.

Governance lesson: Businesses should prohibit the use of pricing tools to implement agreements with competitors concerning prices, customers, output, or other competitive parameters.

Case 3: Trod Ltd. v. Competition and Markets Authority (UK, 2016)

Court: UK Competition Appeal Tribunal proceedings relating to a Competition and Markets Authority infringement decision.

Facts: Trod and GB eye, competing online sellers of posters and frames, agreed not to undercut one another's prices. They used automated repricing software configured to give effect to that arrangement.

Legal issue: Whether automated repricing could be used to implement a price-fixing agreement without removing its unlawful character.

Outcome and principle: Trod admitted the infringement and accepted the CMA's enforcement outcome. The case illustrates that the use of pricing software does not excuse participation in a cartel.

Relevance to algorithmic governance: This case directly illustrates the difference between legitimate automated price competition and algorithmic implementation of an agreement between competitors.

Governance lesson: Compliance programmes should review not only the software itself but also its configuration, commercial objectives, inputs, and instructions supplied by employees.

Case 4: Google LLC and Alphabet Inc. v. European Commission — Google Shopping (Case C-48/22 P, 2024)

Court: Court of Justice of the European Union.

Facts: The dispute concerned Google's treatment of its own comparison-shopping service in general search results and the treatment of competing comparison-shopping services.

Legal issue: Whether a dominant undertaking's conduct involving its search service constituted an abuse of dominance under Article 102 TFEU.

Decision and principle: In September 2024, the Court of Justice dismissed Google's appeal, upholding the General Court's judgment affirming the Commission's infringement decision. The litigation examined the preferential presentation of Google's own service and the potential exclusionary effects on competitors.

Relevance to algorithmic governance: Search ranking is an important form of algorithmic market governance. A dominant platform can influence competitors' visibility and ability to reach customers through the design and operation of its ranking systems.

Governance lesson: Competition authorities should examine whether ranking criteria, preferential treatment, and access conditions unlawfully disadvantage competing services. The mere use of a ranking algorithm, however, is not itself an infringement.

Case 5: United States v. Apple Inc. — e-books price-fixing litigation (2013–2015)

Court: United States District Court for the Southern District of New York; affirmed by the US Court of Appeals for the Second Circuit.

Facts: The litigation concerned Apple's role in a conspiracy with major publishers to raise retail e-book prices through changes to distribution arrangements, including agency agreements and price-parity provisions.

Legal issue: Whether Apple's conduct helped organise and facilitate horizontal price fixing among publishers.

Decision and principle: The courts upheld the finding that Apple had participated in an unlawful price-fixing conspiracy.

Relevance to algorithmic governance: Although the case was not decided as an algorithmic-collusion case, it illustrates the importance of intermediaries and contractual arrangements in facilitating coordination among competitors. Comparable risks can arise when a digital platform or software provider coordinates the commercial decisions of multiple businesses.

Governance lesson: Examine whether a platform's contracts, technical systems, or information-sharing arrangements facilitate coordination between competing suppliers rather than simply providing neutral infrastructure.

Case 6: Rajasthan Cylinders and Containers Ltd. v. Union of India (2018) 13 SCC 472

Court: Supreme Court of India.

Facts: The dispute arose from allegations of cartelisation in the supply of LPG cylinders to public-sector oil marketing companies. The case involved parallel bidding patterns and similarities in the conduct of suppliers.

Legal issue: Whether parallel commercial conduct and similar bidding behaviour were sufficient to establish cartelisation.

Decision and principle: The Supreme Court set aside the Competition Commission of India's findings in the circumstances of the case, emphasising the need to evaluate the evidence in its economic and market context rather than treating parallel conduct as conclusive proof of collusion.

Relevance to algorithmic governance: This principle is highly relevant when competing firms use automated bidding or pricing systems. Similar algorithmic outputs may arise from common market conditions, public information, or independent optimisation rather than an unlawful agreement.

Governance lesson: Authorities should assess the full evidentiary picture, including communications, algorithm design, market conditions, incentives, and conduct that may indicate concertation.

Case 7: Samir Agrawal v. Competition Commission of India (2021) 3 SCC 136

Court: Supreme Court of India.

Facts: The litigation concerned allegations of price coordination involving drivers operating through the Uber platform. The platform's pricing system calculated fares and drivers' remuneration under a centrally organised model.

Legal issue: Whether the platform's pricing mechanism established a horizontal agreement among drivers to fix prices.

Decision and principle: The Supreme Court upheld the CCI's decision not to proceed on the information presented. The Court found that the necessary agreement among drivers had not been established merely because they used a common platform and its pricing mechanism.

Relevance to algorithmic governance: The case illustrates the distinction between a platform centrally setting or calculating prices and independent competitors agreeing to fix prices. The legal analysis depends on the actual relationships, decision-making structure, and applicable statutory requirements.

Governance lesson: Competition authorities should investigate who determines prices, how the system operates, whether competitors exchange sensitive information, and whether evidence establishes the required agreement.

Case 8: Google LLC and Alphabet Inc. v. European Commission — Google AdSense for Search (Case T-334/19, 2024)

Court: General Court of the European Union.

Facts: The dispute concerned contractual restrictions imposed by Google on publishers participating in its search-advertising intermediation services.

Legal issue: Whether the restrictions imposed by a dominant undertaking constituted an abuse of dominance under EU competition law.

Decision and principle: In September 2024, the General Court annulled the Commission's decision in the case. The judgment illustrates the importance of properly establishing the relevant infringement and evaluating the evidence supporting the authority's findings.

Relevance to algorithmic governance: Advertising intermediation systems combine automated matching, targeting, auctions, and contractual access rules. Their competitive effects may depend on both the software and the commercial restrictions surrounding it.

Governance lesson: Effective oversight should investigate algorithmic operations and contractual arrangements together, while requiring competition authorities to prove infringements using the applicable legal standards.

These cases collectively establish an important principle: competition law can address algorithm-enabled conduct through existing legal doctrines, but the specific facts and statutory elements remain decisive. The cases are not all direct rulings on autonomous AI; several provide foundational rules for analysing digital platforms and automated commercial systems.

5. Additional relevant case laws

The following authorities broaden the analysis of market power, intermediary responsibility, and competitive conduct.

Case 9: United States v. Microsoft Corp. (D.C. Cir., 2001)

Court: United States Court of Appeals for the District of Columbia Circuit.

Principle: The case examined Microsoft's conduct concerning its operating-system monopoly and its restrictions affecting browser competition. The judgment addressed exclusionary conduct and the use of control over a dominant technological platform to disadvantage rivals.

Relevance: Algorithmic platforms may similarly control access to users, distribution channels, or complementary services. The case provides a foundation for examining whether technological design choices unlawfully preserve monopoly power.

Case 10: Commission v. Anic Partecipazioni SpA (Case C-49/92 P, 1999)

Court: Court of Justice of the European Communities.

Principle: The judgment addressed the concept of concerted practices and the relationship between coordination and subsequent market conduct.

Relevance: The distinction between independent conduct and concertation is essential when algorithms respond to competitors' prices. Automated similarity alone is not a substitute for the applicable legal proof of coordination.

Case 11: Dole Food Co. v. European Commission (Case C-286/13 P, 2015)

Court: Court of Justice of the European Union.

Principle: The case concerned information exchanges and coordination in the banana market. It clarified aspects of concerted practices and the evidentiary relationship between information exchange and market conduct.

Relevance: Algorithmic systems may process commercially sensitive information at high speed. The case provides useful principles for assessing whether information exchanges reduce strategic uncertainty between competitors.

Case 12: Competition Commission of India v. Steel Authority of India Ltd. (2010) 10 SCC 744

Court: Supreme Court of India.

Principle: The Court clarified important procedural aspects of the Competition Act, 2002, including the nature of the CCI's preliminary decision to direct an investigation.

Relevance: Investigating algorithmic markets may require specialised technical evidence, data analysis, and scrutiny of software configurations. Procedural fairness and proper exercise of statutory powers remain essential throughout the investigation.

6. Application of algorithmic governance under Indian competition law

India's Competition Act, 2002, provides a framework for addressing several algorithmic competition risks. The relevant provisions must be applied to the facts and the legal requirements of each case.

Practical illustrations

Scenario A: Competing retailers use a pricing algorithm.

If retailers independently use software to react to publicly available prices, that fact alone does not establish a cartel. If they agree to use the software to implement common prices, however, Section 3 may become relevant.

Scenario B: A dominant e-commerce platform favours its own products.

If the platform uses its ranking system to disadvantage rival sellers, the CCI would need to assess dominance, the actual conduct, competitive effects, and any legitimate justification under Section 4.

Scenario C: A common software provider serves competing businesses.

If a provider collects competitively sensitive information from competing businesses and uses it to coordinate their pricing, the arrangement may raise concerns under Section 3. The provider's role, knowledge, conduct, and the evidence of coordination would require examination.

Scenario D: AI personalises prices.

Different consumers may receive different offers because of demand, inventory, location, or individual characteristics. Personalised pricing is not automatically unlawful under competition law. It may warrant scrutiny if it forms part of exclusionary or exploitative conduct prohibited by the applicable legal framework.

7. Institutional framework for effective algorithmic governance

An effective competition authority should combine traditional legal investigation with specialised technical capabilities.

1. Algorithmic risk assessment

Identify markets where automated pricing, concentrated platforms, common intermediaries, or data advantages create significant competition risks.

2. Evidence preservation

Examine pricing histories, software versions, model instructions, access logs, business communications, and relevant data exchanges.

3. Technical testing

Use simulations, independent audits, and controlled testing to investigate whether an algorithm facilitates coordination or discriminatory access.

4. Legally grounded enforcement

Connect technical findings to the statutory elements of an infringement, assess counter-evidence, and protect due process.

5. Proportionate remedies

Consider targeted changes to software, information-sharing arrangements, access conditions, monitoring requirements, or other remedies appropriate to the proven harm.

8. Challenges in regulating algorithmic markets

Several difficulties complicate the application of competition law to algorithmic governance.

Opacity: Complex models may make it difficult to determine why a pricing or ranking decision occurred.

Causation: It can be difficult to distinguish algorithmic conduct that causes competitive harm from conduct that merely responds to market conditions.

Tacit coordination: Independent algorithms may converge on similar outcomes without an identifiable agreement.

Evidence and attribution: Authorities must determine which people or businesses designed, instructed, deployed, or knowingly relied on the system.

Innovation concerns: Excessive restrictions may deter beneficial automation, product improvements, and efficient pricing.

Cross-border enforcement: Algorithms, data, software providers, and affected markets may be located in different jurisdictions.

Remedial effectiveness: A remedy must address the source of the competitive harm without unnecessarily disrupting legitimate operations.

These difficulties support stronger technical expertise and more effective evidence-gathering methods, not automatic liability for every algorithmic market outcome.

9. Recommendations for a modern governance model

A balanced model of algorithmic competition governance should include the following measures:

Algorithmic compliance programmes: Require businesses to assess cartel and dominance risks before deploying commercially significant systems.

Independent audits where justified: Prioritise high-risk systems and tailor audit requirements to the suspected or established harm.

Human accountability: Ensure that responsibility for commercial decisions remains identifiable even where automated tools are used.

Data-access safeguards: Prevent the misuse of commercially sensitive information obtained through platforms or shared service providers.

Targeted transparency: Give regulators access to information necessary for enforcement without automatically requiring public disclosure of trade secrets.

Interoperability and contestability: Consider measures that lower unjustified barriers to entry and switching where market conditions and applicable law support them.

Regulatory cooperation: Encourage competition authorities to cooperate with data-protection, consumer-protection, and digital-market regulators within their respective mandates.

Periodic evaluation: Review whether remedies actually restore competitive conditions and whether new technological developments create additional risks.

10. Conclusion

Theories of algorithmic governance provide a framework for understanding how competition law should respond to automated pricing, platform gatekeeping, data-driven market power, algorithmic coordination, and AI-assisted commercial decision-making.

The central lesson from the case law is that the use of an algorithm neither immunises a business from competition law nor independently establishes an infringement. Authorities must examine the underlying agreement or conduct, the market context, the evidence, and the statutory requirements.

The Eturas, Topkins, and Trod matters demonstrate how software can implement unlawful coordination. Google Shopping illustrates how the operation of a dominant platform's ranking system may raise exclusionary concerns. Rajasthan Cylinders and Samir Agrawal highlight the importance of distinguishing parallel or centrally determined conduct from a legally established agreement.

Ultimately, effective algorithmic governance requires a combination of legal accountability, economic evidence, technical competence, procedural fairness, and proportionate remedies. Its objective is not to eliminate algorithms from competitive markets, but to ensure that technological innovation strengthens rather than undermines effective competition.