Algorithmic Intent Doctrine and Liability Attribution in Competition Law

1. Introduction

The Algorithmic Intent Doctrine concerns the assessment of intention, knowledge, foreseeability, and responsibility when firms use algorithms, artificial intelligence (AI), automated pricing tools, recommendation engines, ranking systems, or machine-learning models in ways that may restrict competition.

Traditional competition law frequently examines whether undertakings deliberately entered into an agreement, coordinated commercial conduct, abused a dominant position, or pursued an exclusionary strategy. Algorithmic decision-making complicates this assessment because commercial outcomes may emerge from automated systems without a human employee selecting each individual action.

The central legal question is:

When an algorithm produces an anticompetitive outcome, who should be held liable—the company deploying it, its directors, the software developer, the platform operator, or the algorithm's users?

The answer depends on the applicable jurisdiction, the legal infringement alleged, the conduct of each participant, and the available evidence. An algorithm is generally treated as a tool or system through which legal persons conduct business, rather than as an independent legal person bearing competition-law liability.

The doctrine is best understood as an analytical framework rather than a universally recognized, standalone legal doctrine. Courts and competition authorities generally apply established principles of agreement, concerted practice, unilateral abuse, attribution, knowledge, and corporate responsibility to algorithmically mediated conduct.

2. Meaning and Scope of the Algorithmic Intent Doctrine

The doctrine examines whether an undertaking's use, design, adoption, or continued operation of an algorithm demonstrates legally relevant intention, knowledge, participation, or responsibility for conduct that harms competition.

Its application involves five principal questions:

Knowledge: Did the undertaking know, or have legally relevant reason to know, that the algorithm could facilitate anticompetitive conduct?

Purpose: Was the algorithm designed, configured, or deployed to coordinate prices, exclude rivals, discriminate against competitors, or restrict market access?

Participation: Did the undertaking knowingly participate in an arrangement involving algorithmic coordination?

Foreseeability and control: Could the undertaking reasonably anticipate, monitor, modify, or discontinue the relevant behaviour?

Attribution: Can the conduct of the algorithm, developer, employees, platform, or service provider legally be attributed to a particular undertaking?

These questions must be distinguished from the mere existence of an anticompetitive market outcome. Parallel prices, similar recommendations, or automated refusals do not, by themselves, establish an unlawful agreement or abuse of dominance.

3. Legal Framework Governing Algorithmic Intent and Liability

A. Agreement and concerted practices

Under Article 101 of the Treaty on the Functioning of the European Union (TFEU), Section 1 of the Sherman Act in the United States, and Section 3 of India's Competition Act, 2002, liability may arise where the statutory requirements for agreements, coordination, or anticompetitive arrangements are established.

Algorithmic systems may facilitate coordination by:

Collecting and transmitting competitors' commercially sensitive information.

Implementing common pricing instructions.

Monitoring compliance with an agreement.

Automatically adjusting prices in response to rivals.

Making deviations from agreed prices more detectable.

Enforcing contractual or technical restrictions on independent commercial decisions.

However, independently developed algorithms responding to common market conditions may produce similar results without any unlawful agreement. Similarity of outcomes must therefore be assessed alongside evidence of communication, knowledge, participation, and other legally relevant circumstances.

B. Abuse of dominance and exclusionary intent

Under Article 102 TFEU, Section 4 of India's Competition Act, and Section 2 of the Sherman Act, algorithmic systems may be relevant to allegations involving:

Self-preferencing and discriminatory rankings.

Exclusionary access restrictions.

Predatory pricing or strategic price responses.

Tying and bundling through digital interfaces.

Restrictions on interoperability.

Exploitative or discriminatory trading conditions, where prohibited by the applicable law.

In unilateral-conduct cases, proof of an agreement with competitors is generally unnecessary. The principal questions concern dominance or market power, the undertaking's conduct, the applicable legal test, and whether the conduct constitutes an abuse or unlawful monopolization.

C. Corporate attribution and responsibility

An undertaking cannot ordinarily avoid competition-law responsibility merely by claiming that a computer independently made the relevant decision.

Where a firm selects the algorithm's objectives, supplies the data, defines its constraints, authorizes its deployment, or benefits from its operation, those facts may support attribution of the resulting commercial conduct to the firm. The precise legal consequences depend on the jurisdiction and infringement alleged.

Conversely, a software developer does not automatically become liable merely because its product is used by a customer to engage in unlawful conduct. Liability depends on the developer's own participation, knowledge where legally relevant, contribution to the infringement, and the applicable legal rules.

4. Important Case Laws on Algorithmic Intent and Liability Attribution

The following cases establish principles relevant to algorithmic competition enforcement. Some directly concern computerized pricing or digital algorithms, while others provide analogous principles on cartel facilitation, corporate responsibility, and exclusionary conduct.

Case 1: Eturas UAB and Others v Lietuvos Respublikos konkurencijos taryba (2016)

Citation: Case C-74/14, Court of Justice of the European Union.

Facts: Several travel agencies used a common online booking system. The platform administrator circulated a message concerning discount restrictions, and the system was technically modified to make larger discounts more difficult to offer.

Legal issues:

Whether participating agencies could be presumed to know about the restriction.

Whether failure to oppose the measure indicated tacit acceptance.

Whether the platform operator and participating agencies could be held liable for a concerted practice.

Decision and principles: The Court held that participation could not automatically be inferred merely from the sending of a message or the agencies' use of the system. Awareness and participation had to be assessed through the evidence, subject to the applicable rules on proof and rebuttal. An undertaking could rebut a presumption of knowledge through relevant evidence, including public distancing from the practice or reporting it to the authorities.

Relevance to algorithmic intent: This is a particularly relevant case because it illustrates the distinction between an automated restriction and the knowledge or participation of individual businesses using the system. A technical change affecting several competitors does not, by itself, establish that each competitor agreed to coordinate.

Legal principle: Algorithmic implementation may provide evidence of coordination, but liability requires proof satisfying the applicable legal standard.

EUR-Lex

+1

Case 2: AC-Treuhand AG v European Commission (2015)

Citation: Case C-194/14 P, Court of Justice of the European Union.

Facts: AC-Treuhand, a consultancy firm, assisted manufacturers involved in chemical-stabilizer cartels. Its activities included organizing meetings, collecting commercially sensitive information, and facilitating cartel operations, although it was not itself a producer in the cartelized markets.

Legal issues:

Whether a third party outside the relevant product market could be liable for cartel participation.

Whether facilitation could constitute participation in an infringement of competition law.

Decision: The Court upheld the finding that a facilitator could be liable under EU competition law despite not competing in the affected markets itself.

Relevance to algorithmic intent: A third-party pricing-software provider, data intermediary, or platform operator may potentially incur liability if its conduct amounts to knowing and intentional participation in a cartel. Merely supplying general-purpose software, however, does not automatically establish such participation.

Legal principle: Liability may extend beyond the competing firms to a third party that actively contributes to an anticompetitive arrangement with the requisite knowledge and participation.

Publications Office of the EU

+1

Case 3: United States v. David Topkins (2015)

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

Facts: The case concerned an agreement among sellers of posters and related products sold through an online marketplace. The participants used pricing algorithms to implement coordinated pricing arrangements.

Legal issues:

Whether algorithmic pricing could implement an unlawful price-fixing agreement.

Whether using software changed the legal character of the underlying agreement.

Whether an individual involved in the arrangement could be held personally accountable.

Outcome: David Topkins pleaded guilty to participating in a price-fixing conspiracy involving online poster sales.

Relevance to algorithmic intent: The case demonstrates that an algorithm can be an instrument for implementing an agreement rather than a substitute for proof of the agreement itself. Automated repricing does not remove liability where the underlying conduct constitutes unlawful coordination.

Legal principle: The use of algorithms does not immunize an otherwise unlawful price-fixing arrangement from antitrust enforcement.

Case 4: Trod Ltd v Competition and Markets Authority (2016)

Citation: UK Competition Appeal Tribunal, Case 1266/1/12/16.

Facts: Trod Ltd and another online retailer agreed not to undercut each other's prices for certain products sold through Amazon's UK marketplace. Automated repricing software was used in the relevant commercial activities.

Legal issues:

Whether online resale price coordination infringed competition law.

Whether automated pricing mechanisms affected responsibility for the underlying arrangement.

Whether the parties could rely on the technical operation of their pricing systems as a defence.

Outcome: The case followed enforcement action by the Competition and Markets Authority concerning the parties' agreement not to undercut one another.

Relevance to algorithmic intent: The case illustrates the importance of distinguishing an algorithm that independently responds to market conditions from an algorithm configured or used to implement an existing agreement between competitors.

Legal principle: Automated price-setting is not a defence to liability for an established anticompetitive agreement.

Case 5: Google LLC and Alphabet Inc. v European Commission (Google Shopping) (2024)

Citation: Case C-48/22 P, Court of Justice of the European Union, Grand Chamber.

Facts: Google was found to have favoured its own comparison-shopping service in general search results while competing services were disadvantaged by the way results were displayed and ranked.

Legal issues:

Whether differential treatment within an algorithmically operated platform could constitute an abuse of dominance.

How exclusionary effects and the relationship between conduct and market outcomes should be assessed.

Whether the conduct could be justified by legitimate competitive considerations.

Decision: On 10 September 2024, the Court of Justice dismissed Google's appeal, leaving the relevant finding of abuse in place.

Relevance to algorithmic intent: The case shows that competition-law analysis may focus on how a dominant undertaking designs and applies ranking systems, the competitive conditions it creates, and the capacity of the conduct to foreclose rivals. It does not establish that proof of a subjective intention to exclude is universally required for algorithmic abuse.

Legal principle: Algorithmic ranking and differential treatment may be assessed as part of a broader abuse-of-dominance inquiry, even without proof of a cartel agreement.

EUR-Lex

+1

Case 6: United States v. Apple Inc. (2013)

Citation: 952 F. Supp. 2d 638 (S.D.N.Y. 2013), affirmed in relevant part, 791 F.3d 290 (2d Cir. 2015).

Facts: The United States challenged arrangements involving Apple and publishers concerning the pricing of electronic books. The case concerned coordinated commercial arrangements and the use of an agency pricing model.

Legal issues:

Whether the parties' conduct formed part of an unlawful conspiracy.

Whether contractual and commercial structures could facilitate coordinated pricing.

How circumstantial evidence could establish participation in an anticompetitive arrangement.

Decision: The district court found that Apple had participated in a conspiracy to raise e-book prices, and the Second Circuit affirmed the relevant judgment.

Relevance to algorithmic intent: Although not an algorithmic-pricing case, it provides a useful analogy for assessing the role of an intermediary that structures commercial relationships and facilitates coordination among market participants.

Legal principle: Competition authorities may examine the substance of commercial arrangements, the role of intermediaries, and the evidence of coordination rather than treating formal contractual structures as conclusive.

Case 7: T-Mobile Netherlands BV and Others (2009)

Citation: Case C-8/08, Court of Justice of the European Union.

Facts: The case involved exchanges of commercially sensitive information among mobile telecommunications operators.

Legal issues:

Whether a single information exchange could facilitate a concerted practice.

Whether direct evidence of a formal agreement was necessary.

How knowledge of competitors' intended market conduct should be evaluated.

Decision: The Court explained that a single meeting or exchange may be sufficient to establish a concerted practice where the relevant legal requirements are met, depending on the circumstances and evidence.

Relevance to algorithmic intent: Shared pricing dashboards, competitor-data feeds, and common optimization systems can raise comparable concerns when they communicate competitively sensitive information and reduce uncertainty about competitors' future conduct.

Legal principle: The relevant inquiry concerns the nature and purpose of the information exchange and its relationship to coordinated market conduct, rather than the number of communications alone.

Case 8: Intel Corp. v European Commission (2024)

Citation: Case C-240/22 P, Court of Justice of the European Union, Grand Chamber, 24 October 2024.

Facts: The case concerned rebates offered by Intel and allegations of abuse of its dominant position in the market for x86 central processing units. It followed extensive proceedings regarding the Commission's assessment of exclusionary conduct.

Legal issues:

How exclusionary conduct by a dominant undertaking should be assessed.

The evidentiary requirements for evaluating whether conduct is capable of restricting competition.

The treatment of economic evidence in abuse-of-dominance proceedings.

Decision: The Court dismissed the Commission's appeal concerning the General Court's judgment, which had annulled the relevant decision on the contested rebate assessment.

Relevance to algorithmic intent: The case is relevant by analogy where an undertaking uses automated rebates, targeted pricing, ranking rules, or incentive systems. It illustrates why a competition authority must apply the appropriate substantive test and evaluate relevant evidence rather than treating the mere presence of an exclusionary-looking strategy as sufficient.

Legal principle: Liability for unilateral conduct must be established under the governing legal test, with appropriate attention to the evidence and competitive effects.

5. Distinguishing Intent, Knowledge, Foreseeability, and Outcome

The algorithmic intent doctrine requires careful differentiation between concepts that may overlap but are not legally identical.

An important distinction is that subjective intent is not a universal prerequisite for competition-law liability. Certain forms of cartel conduct may be established through evidence of agreement and participation, while abuse-of-dominance cases may focus on the nature of the conduct and its exclusionary capacity or effects.

Similarly, an algorithm's unexpected behaviour does not automatically absolve an undertaking. Its significance depends on the applicable legal standard, the facts, and the firm's response after discovering the problem.

6. Attribution of Liability Among Different Actors

Algorithmic systems often involve several independent participants. Each should be assessed according to its own conduct rather than assigning collective responsibility merely because the participants contributed to the same technical system.

Liability attribution framework

Algorithmic system or platform

Identify the relevant commercial decision and technical mechanism

Deploying firm

Objectives, configuration, adoption, monitoring, and commercial use

Software provider

Design, customization, knowledge, and contribution to any arrangement

Participating competitors

Communications, information exchange, agreement, or coordinated conduct

Employees and managers

Authorization, implementation, knowledge, and individual involvement

The final legal assessment identifies the relevant infringement, the evidence connecting each actor to it, and the applicable rules for corporate or individual liability.

A. Liability of the deploying undertaking

The company that uses the algorithm may be liable where the algorithm implements its commercial policy or forms part of conduct attributable to it.

Relevant evidence may include:

Internal instructions concerning pricing, ranking, or exclusion.

Approval of algorithmic objectives and constraints.

Monitoring reports and responses to identified risks.

Decisions to retain a system after discovering unlawful conduct.

Communications with competitors or a common software provider.

B. Liability of the software developer

A developer may face liability where its own conduct satisfies the applicable elements of an infringement, such as knowingly facilitating a cartel or actively participating in coordination.

The provision of ordinary software, hosting, or technical support is not, without more, sufficient to establish cartel participation. The developer's role, knowledge, contribution, and contractual and operational activities must be examined.

C. Liability of participating competitors

Competitors may be liable where they knowingly participate in an agreement or concerted practice that infringes competition law. A common pricing tool may provide evidence of the mechanism through which coordination occurred, but the mere use of the same software does not establish an agreement.

D. Individual liability of directors and employees

Individual liability varies significantly by jurisdiction. Directors and employees may face personal consequences where the applicable law permits individual sanctions and the evidence establishes their involvement. Corporate liability should not automatically be equated with personal criminal liability.

7. Application Under Indian Competition Law

In India, algorithmic intent and liability attribution can be examined primarily through the Competition Act, 2002, as amended.

A. Section 3: Anti-competitive agreements

Section 3 prohibits specified agreements that cause or are likely to cause an appreciable adverse effect on competition in India. Depending on the circumstances, algorithmic arrangements may be relevant to:

Price fixing.

Bid rigging.

Market allocation.

Restrictions on production or supply.

Vertical restraints, including resale price maintenance, exclusive dealing, and refusal to deal.

Where competitors use a shared algorithm to implement a price-fixing agreement, the legal analysis should focus on the underlying agreement, participation, evidence, and statutory requirements. Parallel algorithmic pricing alone does not necessarily establish an infringement.

B. Section 4: Abuse of dominant position

Section 4 may be relevant where a dominant digital platform uses algorithms to engage in prohibited conduct, including unfair or discriminatory conditions, denial of market access, or other conduct covered by the provision.

Examples include:

Ranking a platform's own products more favourably than competing products.

Restricting access to commercially necessary platform infrastructure.

Using algorithmic rules to foreclose rival service providers.

Imposing discriminatory conditions through automated allocation systems.

Dominance alone is not unlawful. The conduct must satisfy the requirements for abuse under Section 4.

C. Sections 26 and 27: Investigation and remedies

Section 26 governs the process for directing investigations and related procedural steps. Section 27 provides for orders and consequences following findings of specified contraventions.

Digital evidence relevant to an investigation may include source-code documentation, algorithmic logs, pricing histories, internal communications, configuration changes, and records of access to shared data. Such evidence must be assessed in accordance with applicable procedural and evidentiary rules.

D. Section 48: Offences by companies

Section 48 addresses contraventions by companies and the circumstances in which persons responsible for the conduct of the business, as well as persons whose consent, connivance, or neglect is established, may be liable under the statutory framework.

The section should not be interpreted as automatically imposing personal liability on every director or employee merely because a company used an algorithm. The statutory conditions and available defences must be examined.

8. Evidence Used to Establish Algorithmic Intent

Competition authorities and courts may need to combine technical evidence with conventional documentary, economic, and testimonial evidence.

No single category is necessarily conclusive. For example, source code that permits rapid price matching does not itself prove an unlawful agreement. Likewise, an internal document discussing competitor monitoring may have legitimate procompetitive explanations.

A sound assessment examines the totality of the evidence, considers plausible alternative explanations, and applies the appropriate standard of proof.

9. Defences and Limitations

Several considerations may limit or defeat a finding of algorithmic competition-law liability.

Independent decision-making: Similar algorithmic outcomes may result from common demand conditions, public market information, or independently selected commercial strategies.

Lack of participation: A business may use common software without knowing about or participating in a competitor's unlawful agreement.

Legitimate commercial purpose: Algorithms can improve inventory management, forecasting, logistics, customer matching, and pricing efficiency.

Objective justification: In abuse-of-dominance cases, the undertaking may argue that the challenged conduct is objectively justified or produces verifiable efficiencies, where such considerations are legally available.

Insufficient evidence: An authority must establish the relevant infringement and cannot simply equate an unexplained outcome with unlawful intention.

Jurisdictional differences: The relevance of intent, the burden of proof, the treatment of facilitators, and the availability of individual sanctions vary among legal systems.

10. Practical Hypothetical

Assume that three competing online retailers use a shared pricing platform. The platform operator introduces a feature that recommends a common minimum price based on commercially sensitive information supplied by all three retailers.

The retailers subsequently charge similar prices.

Three possible situations illustrate how liability may differ.

Situation A: Independent price optimization. Each retailer independently supplies its own inventory and publicly available market prices. The software recommends similar prices because market conditions are similar. Without additional evidence, similarity alone may be insufficient to establish unlawful coordination.

Situation B: Knowing coordination. The retailers agree to maintain a minimum price, provide confidential future pricing information to the platform, and configure the software to implement their agreement. The algorithm is used as a mechanism for carrying out the cartel. The retailers may be liable if the applicable legal elements are established.

Situation C: Platform-facilitated coordination. The platform operator knowingly designs and operates a system to communicate confidential competitor information and enforce the agreed minimum price. The operator's liability would depend on whether its own conduct satisfies the jurisdiction's rules governing cartel facilitation and participation.

The key distinction is not simply whether an algorithm produced similar prices. It is whether the evidence establishes the relevant agreement, participation, abuse, or other prohibited conduct attributable to each party.

11. Regulatory and Compliance Measures

Businesses using algorithmic decision systems can reduce competition-law risks through governance measures that preserve legitimate commercial autonomy.

Independent pricing controls: Ensure that each undertaking independently determines its commercial pricing policy.

Competitor-data restrictions: Avoid unauthorized exchanges of confidential future prices, output plans, margins, or strategic intentions.

Algorithmic impact assessments: Assess whether system design or deployment could facilitate coordination or exclusion.

Audit trails: Preserve relevant records of system configuration, model updates, inputs, and consequential decisions.

Human oversight: Establish appropriate review and override mechanisms for high-impact commercial decisions.

Vendor due diligence: Examine whether shared software transmits sensitive information between competing users or embeds coordinated pricing instructions.

Incident response: Investigate and correct identified anticompetitive risks promptly, while obtaining appropriate legal advice.

Training and accountability: Ensure that commercial, legal, engineering, and data teams understand the competition-law implications of algorithmic systems.

These controls do not guarantee immunity from liability. Their effectiveness depends on implementation, the conduct involved, and the applicable law.

12. Critical Analysis

The principal difficulty in algorithmic competition cases is that automated systems can separate the commercial decision-maker from the immediate mechanism that produces the outcome.

Traditional evidence may include meetings, written agreements, price lists, or direct communications. Algorithmic coordination may instead involve shared data feeds, automated responses, common optimization objectives, and repeated technical interactions.

Three analytical challenges follow.

First, intent must not be confused with outcome. An algorithm that raises prices may be operating independently, implementing an unlawful agreement, or engaging in conduct that raises different competition concerns. The outcome alone does not resolve the legal classification.

Second, technical autonomy must not become a blanket defence. A firm cannot necessarily avoid responsibility by delegating a commercial decision to software or describing the resulting conduct as machine-generated.

Third, liability must remain individualized. A developer, platform operator, customer, and employee may play different roles. Attribution should depend on evidence and the governing legal rules rather than the mere fact that each actor participated in the same technological ecosystem.

The appropriate approach is therefore neither to presume unlawful intent whenever algorithms produce parallel outcomes nor to exempt algorithmic conduct from conventional competition-law scrutiny.

13. Conclusion

The Algorithmic Intent Doctrine provides a useful framework for examining the relationship between automated commercial decisions, knowledge, participation, and legal responsibility.

The cases discussed above demonstrate several complementary principles:

Eturas addresses knowledge, tacit participation, and evidentiary presumptions in a computerized booking system.

AC-Treuhand establishes that a qualifying cartel facilitator may be liable even when it does not compete in the affected market.

Topkins and Trod illustrate that software does not shield participants from liability for established price-fixing arrangements.

Google Shopping demonstrates the relevance of algorithmic ranking and differential treatment to abuse-of-dominance analysis.

Apple illustrates how intermediaries and commercial arrangements can contribute to coordination.

T-Mobile Netherlands addresses information exchange and concerted practices.

Intel illustrates the need to apply the correct legal test and evaluate the evidence in exclusionary-conduct cases.