Competition Law and Collective Intelligence Systems and Antitrust

1. Introduction

Collective Intelligence Systems (CIS) are digital systems in which information, predictions, behavioural data, machine learning, algorithms, artificial intelligence, or networked decision-making tools combine information generated by multiple participants and use it to produce collective recommendations or decisions.

Examples include:

• AI-powered pricing systems;
• common data and analytics platforms;
• algorithmic procurement systems;
• industry-wide forecasting platforms;
• shared demand-prediction systems;
• digital marketplaces;
• AI-based recommendation engines;
• common logistics and capacity-management systems;
• cloud-based pricing software;
• collective credit or risk-scoring systems; and
• AI systems trained using competitors' commercially sensitive information.

Collective intelligence itself is not unlawful. It may produce substantial efficiencies, better forecasting, lower transaction costs and innovation. The antitrust problem arises when the system becomes a mechanism through which independent competitors coordinate prices, output, customers, markets, investment, innovation or competitively sensitive information.

The central question is therefore:

When does legitimate collective intelligence become a mechanism for collective restriction of competition?

Traditional competition law generally does not require the unlawful coordination to occur through face-to-face meetings. An agreement may be implemented through software, a common intermediary, data-sharing arrangement or automated algorithm.

2. Legal Framework

Collective intelligence systems can engage several areas of competition law.

A. Agreements and concerted practices

The principal issue is whether participating undertakings have:

• expressly agreed to coordinate;
• exchanged commercially sensitive information;
• knowingly participated in a common system;
• accepted common pricing parameters;
• delegated competitive decisions to a common algorithm; or
• deliberately adopted a system that predictably coordinates their behaviour.

Under EU law, Article 101 TFEU is particularly important.

In the United States, Section 1 of the Sherman Act addresses agreements restraining trade.

In India, Section 3 of the Competition Act, 2002 addresses agreements having or likely to have an appreciable adverse effect on competition.

3. Why Collective Intelligence Creates Antitrust Risks

3.1 Information aggregation

A CIS may collect:

• prices;
• discounts;
• inventory;
• capacity;
• production levels;
• customer information;
• future business plans;
• wages;
• demand forecasts;
• strategic investment information.

Information that would ordinarily remain confidential can become available through a common system.

The competition concern is particularly serious where the information is:

1. commercially sensitive;
2. non-public;
3. granular;
4. current or forward-looking; and
5. supplied by competing undertakings.

3.2 Algorithmic coordination

Suppose ten competing retailers independently provide information to an AI platform.

The platform analyses their data and recommends:

"Retailers should charge approximately ₹1,000."

If all ten retailers independently follow the recommendation, the system may reduce price competition.

The legal difficulty is determining whether this is:

• genuinely independent parallel conduct;
• conscious coordination;
• an information-exchange arrangement;
• a hub-and-spoke arrangement; or
• an unlawful agreement implemented by technology.

4. Principal Models of Collective Intelligence Antitrust Risk

Model 1: Messenger Model

Competitors first make an unlawful agreement and subsequently use software to implement it.

Example

A group of online sellers agree:

"We will not sell below ₹500."

They then program their pricing systems to maintain ₹500.

The algorithm does not create the cartel. It merely implements the cartel.

This is the easiest case for competition authorities because the underlying agreement can establish liability.

5. Case Law

Case 1 — United States v. Topkins

United States v. David Topkins (2015) is one of the most important early algorithmic-pricing cases.

Online sellers of posters allegedly agreed to coordinate prices on Amazon Marketplace. They agreed to use specific pricing algorithms to implement their pricing arrangement.

The Department of Justice treated the algorithm as the technological mechanism used to implement an ordinary price-fixing conspiracy.

Principle

Technology does not immunise an otherwise unlawful cartel.

If competitors first agree to fix prices and then use an algorithm to implement that agreement, conventional cartel principles continue to apply.

Importance for CIS

Topkins demonstrates the Messenger Model:

Human agreement → algorithm → coordinated market behaviour

rather than:

algorithm → agreement.

6. Case 2 — Eturas UAB and Others v Lietuvos Respublikos konkurencijos taryba

Case C-74/14, Court of Justice of the European Union (2016)

This is particularly significant for collective digital systems.

Travel agencies used a common online booking system operated by Eturas. The system administrator communicated a message concerning a restriction on discounts available to customers.

The CJEU considered whether knowledge of the common system and failure to distance oneself could establish participation in a concerted practice.

The Court recognised that a common digital system can provide the mechanism through which coordination occurs, but also emphasised the evidentiary requirements and the importance of rebutting presumptions appropriately.

Principle

A digital system can be relevant evidence of coordination, but mere receipt of a system message is not automatically sufficient to establish liability.

CIS significance

Eturas is important because it demonstrates the potential movement from:

common digital infrastructure → common information → coordinated behaviour.

It also shows why evidence concerning actual awareness and participation matters.

7. Case 3 — Samir Agarwal v. Competition Commission of India

Supreme Court of India, 2020

This case concerned allegations involving Ola and Uber and the use of algorithmic pricing.

The allegation was essentially that drivers were unable to independently determine prices because the platforms' algorithms determined fares, potentially facilitating coordination among drivers.

The CCI did not find a prima facie agreement establishing price-fixing and closed the matter; the matter subsequently reached the appellate and Supreme Court proceedings.

Principle

The mere existence of an algorithmically determined price does not automatically establish an antitrust agreement.

Competition authorities must still examine whether there is:

• an agreement;
• arrangement;
• understanding;
• concerted practice; or
• another legally sufficient mechanism of coordination.

Importance for India

The case is especially relevant to platform-mediated collective intelligence, where independent participants use a common algorithm.

8. Case 4 — United States v. RealPage

United States v. RealPage, Inc.

RealPage involved algorithmic pricing in the residential rental market.

The U.S. Department of Justice alleged that competing landlords supplied RealPage with non-public information concerning rents and other commercially sensitive information, which was then used by algorithmic pricing software to generate pricing recommendations. The DOJ characterised the alleged conduct as involving Sections 1 and 2 of the Sherman Act.

The government's allegations included a potential hub-and-spoke structure, with RealPage functioning as the technological hub through which competitors' information could contribute to coordinated pricing recommendations.

Principle

A competitor does not necessarily escape antitrust scrutiny merely because:

"The computer, rather than a human employee, made the recommendation."

The relevant question is whether the underlying arrangement facilitates unlawful coordination.

CIS significance

RealPage represents a more sophisticated model:

Competitors → common data pool → algorithm → recommendations → competitors' pricing decisions

This is much closer to a genuine collective-intelligence system.

9. Case 5 — Cornish-Adebiyi v. Caesars Entertainment

This litigation concerns allegations involving hotel-room pricing algorithms.

In 2024, the FTC and DOJ filed a statement of interest addressing the use of algorithmic pricing systems in the hotel industry. The agencies stated that companies cannot avoid antitrust law by using an algorithm to perform conduct that would be unlawful if performed by humans.

Principle

The relevant legal inquiry focuses on the competitive substance of the arrangement, rather than whether the coordination is performed manually or automatically.

CIS significance

If multiple hotels use a common pricing system that receives competitively sensitive information from participating hotels and produces coordinated pricing recommendations, the system may raise:

• information-exchange concerns;
• hub-and-spoke concerns;
• price-fixing concerns; and
• facilitation concerns.

10. Case 6 — T-Mobile Netherlands BV v Raad van bestuur van de Nederlandse Mededingingsautoriteit

CJEU, Case C-8/08

This is a foundational EU information-exchange case.

Mobile telecommunications operators exchanged information concerning dealer remuneration.

The CJEU held that an exchange of information capable of reducing uncertainty between competitors can constitute a restriction of competition when it facilitates coordination.

Principle

Competition law protects the independence of competitive decision-making.

A collective intelligence platform can therefore create antitrust risk even before competitors expressly agree:

"We will charge the same price."

If the system substantially reduces strategic uncertainty between competitors, the information exchange itself may become problematic.

11. Case 7 — AC-Treuhand AG v European Commission

CJEU, Cases C-194/14 P and related jurisprudence

AC-Treuhand is important for the liability of third parties that facilitate cartel activity.

The case concerned a consultancy that played a role in facilitating cartel arrangements.

Principle

Competition-law responsibility can extend beyond the traditional manufacturer-versus-manufacturer cartel where a third party knowingly contributes to the implementation of an anticompetitive arrangement.

CIS significance

This becomes particularly important where the collective intelligence provider is:

• an AI company;
• data intermediary;
• cloud provider;
• pricing-software company;
• industry association;
• marketplace;
• analytics provider; or
• common data platform.

A technology provider may therefore face scrutiny where its role goes beyond neutral technological infrastructure and involves facilitating coordination.

12. Case 8 — United States v. Apple Inc. — E-books

The Apple e-books litigation provides an important illustration of coordinated conduct involving a technological platform and multiple market participants.

The Supreme Court considered whether Apple had facilitated coordination among publishers concerning e-book pricing.

Principle

A platform or intermediary can become competition-law relevant when it facilitates coordinated conduct among otherwise independent competitors.

CIS significance

Modern AI and collective intelligence platforms can similarly become coordination hubs if they allow competitors to align:

• prices;
• commissions;
• contractual terms;
• distribution;
• output;
• customers; or
• strategic decisions.

13. Collective Intelligence and Hub-and-Spoke Cartels

One of the most important antitrust structures is:

Traditional hub-and-spoke

Competitor A
↘
Platform / Hub
↗
Competitor B

The platform receives information from A and B and potentially communicates information or recommendations back to them.

AI-enabled hub-and-spoke

Retailer A ─┐
Retailer B ─┼→ AI/Data Platform → Pricing Recommendation
Retailer C ─┘

The algorithm becomes the common mechanism through which competitive information is processed.

The critical issue is whether participating firms have made a sufficiently established commitment to a common anticompetitive scheme.

RealPage illustrates why this structure is receiving increasing attention in contemporary antitrust enforcement.

14. Collective Intelligence and Tacit Collusion

A more difficult question arises when there is no express agreement.

Imagine competing firms independently deploy AI systems that:

1. monitor competitors' prices;
2. predict competitors' responses;
3. punish price reductions;
4. reward price increases;
5. rapidly adjust prices; and
6. learn from competitors' conduct.

Prices may converge without any employee ever communicating with another competitor.

This is sometimes described as algorithmic or tacit collusion.

The distinction is crucial.

Mere parallel conduct

A firm independently chooses the same price as its competitors.

This does not automatically establish an antitrust agreement.

Coordinated conduct

Firms use a common system, exchange sensitive information, adopt common restrictions, or knowingly participate in a system designed to coordinate their competitive behaviour.

This creates substantially greater antitrust risk.

15. Collective Intelligence and Information Exchange

Information exchanged through CIS can be classified according to competitive sensitivity.

The more current, individualised and forward-looking the information, the greater the potential competition concern.

16. Collective Intelligence and Market Dominance

CIS can also generate Article 102 / Section 2 / Indian dominance concerns.

A dominant AI platform may control:

• essential datasets;
• industry benchmarks;
• market forecasts;
• recommendation systems;
• interoperability standards;
• API access;
• training data;
• customer behavioural information.

A dominant platform might then:

1. deny competitors access to essential data;
2. provide inferior data to rivals;
3. favour its own downstream services;
4. discriminate between users;
5. bundle intelligence services with other products;
6. impose exclusivity;
7. prevent interoperability; or
8. use rivals' information to compete against them.

Thus, collective intelligence creates both cartel risks and unilateral-conduct risks.

17. Data Pooling and Antitrust

Data pooling is not inherently unlawful.

It can create significant efficiencies, such as:

• fraud detection;
• cybersecurity;
• medical research;
• supply-chain optimisation;
• environmental monitoring;
• logistics;
• demand forecasting;
• financial risk assessment.

However, safeguards become important where competitors contribute data.

Safer architecture

Competitors → anonymisation → aggregation → independent statistical output

rather than:

Competitors → identifiable real-time data → common platform → individualised pricing recommendations

The second structure creates considerably greater coordination risks.

18. Collective Intelligence in Procurement

Suppose competing construction companies participate in an AI procurement platform.

The platform knows:

• each firm's reservation price;
• available capacity;
• labour costs;
• expected bids;
• material costs.

If the platform generates recommendations enabling competitors to avoid underbidding each other, the system may undermine competitive tendering.

Possible concerns include:

• bid coordination;
• market allocation;
• bid rotation;
• exchange of competitively sensitive information;
• common pricing formulas.

19. Collective Intelligence in Employment Markets

The same problem can occur in labour markets.

Suppose competing employers jointly use an AI system that collects:

• employee salaries;
• future salary increases;
• hiring plans;
• bonus structures;
• workforce reductions.

The system could potentially reduce competition for labour.

The antitrust analysis therefore extends beyond product prices to wages and employment conditions.

20. Collective Intelligence and AI Training Data

A particularly modern problem concerns AI training datasets.

Several competing firms may collectively contribute:

• customer data;
• transaction information;
• pricing information;
• product-development information;
• proprietary research.

The resulting AI model may become more powerful than any individual participant's model.

Competition law must therefore distinguish between:

Legitimate collaborative innovation

Competitors jointly develop a technology that would be too expensive to develop independently.

and

Anticompetitive coordination

Competitors use the collaborative system to:

• coordinate prices;
• restrict output;
• divide markets;
• exchange strategic information; or
• exclude competitors.

21. Collective Intelligence and Merger Control

CIS may also affect merger analysis.

A merger between:

• a major AI provider;
• a data aggregator;
• a marketplace; and
• an industry analytics platform

could produce substantial concentration of:

• data;
• computational resources;
• customers;
• prediction capabilities;
• infrastructure;
• distribution channels.

Competition authorities may therefore examine whether the transaction creates:

• data concentration;
• foreclosure;
• interoperability problems;
• access restrictions;
• vertical leverage;
• network effects; or
• increased barriers to entry.

22. Collective Intelligence and Network Effects

CIS frequently has strong network effects.

More participants produce more data.

More data produces better predictions.

Better predictions attract more users.

More users generate still more data.

This creates a feedback loop:

Users → Data → Intelligence → Better service → More users → More data

The resulting market structure can produce substantial competitive advantages.

The antitrust question is whether those advantages arise from legitimate innovation or are reinforced through exclusionary conduct.

23. Collective Intelligence as a Potential Essential Facility

In some circumstances, an extremely important collective intelligence platform may become an infrastructure upon which downstream competitors depend.

Potential issues include:

• refusal to provide access;
• discriminatory access;
• excessive access charges;
• interoperability restrictions;
• technical degradation;
• discriminatory APIs;
• exclusionary licensing.

However, the traditional legal requirements for refusal-to-deal or essential-facility theories remain relevant; mere usefulness of a dataset or AI system does not automatically make it an essential facility.

24. Compliance Framework for Collective Intelligence Systems

Businesses using CIS should implement a competition-law governance framework.

1. Data classification

Classify data as:

• public;
• historical;
• aggregated;
• anonymised;
• commercially sensitive;
• forward-looking.

2. Access controls

Competitors should not have unrestricted access to one another's sensitive information.

3. Algorithm governance

Maintain records of:

• algorithm design;
• input data;
• pricing variables;
• optimisation objectives;
• system changes;
• output logic.

4. Independent decision-making

Participants should retain genuine authority to make their own competitive decisions.

5. No competitor-specific recommendations

Particular care should be taken where an AI provider recommends prices using identifiable competitors' confidential information.

6. Auditability

The system should be capable of explaining:

• what data it used;
• how data was aggregated;
• what variables affected recommendations;
• whether competitors' information was incorporated.

7. Competition-law review

High-risk CIS applications should undergo competition-law review before deployment.

25. Evidentiary Problems

Collective intelligence systems create a major evidentiary challenge.

Traditional cartel evidence includes:

• emails;
• meetings;
• telephone calls;
• agreements;
• written instructions.

AI systems may instead leave:

• source code;
• API logs;
• model outputs;
• data-access records;
• model-training records;
• system prompts;
• configuration files;
• audit logs;
• server communications.

Therefore, competition investigations increasingly require technical evidence in addition to traditional documentary evidence.

26. Six Core Legal Questions for Courts and Competition Authorities

When analysing a CIS, authorities can ask:

Question 1

Who supplied the data?

Question 2

Was the data competitively sensitive?

Question 3

Who controlled the system?

Question 4

Did competitors knowingly participate in a common mechanism?

Question 5

Did the system reduce strategic uncertainty between competitors?

Question 6

Did the system actually facilitate coordinated competitive behaviour?

These questions help distinguish legitimate technological cooperation from unlawful coordination.

27. Case-Law Principles Compared

28. Indian Competition-Law Perspective

For India, the principal provision is Section 3 of the Competition Act, 2002.

Collective intelligence systems may potentially raise concerns involving:

Section 3(3)

Where competing enterprises coordinate:

• prices;
• output;
• markets;
• customers;
• bids; or
• other competitive parameters.

Section 3(1)

Even arrangements outside classic cartel categories may be scrutinised where they cause or are likely to cause an appreciable adverse effect on competition.

Section 4

A dominant CIS operator could potentially face issues involving:

• discriminatory access;
• exclusion;
• tying;
• refusal to deal;
• leveraging;
• discriminatory conditions; or
• exploitative conduct.

The Samir Agarwal litigation is particularly useful for understanding the difficulty of proving an agreement in algorithm-mediated markets.

29. Distinguishing Legitimate Collective Intelligence from Antitrust Violations

The table is not a substitute for case-specific legal analysis; the same technology may be lawful in one context and problematic in another.

30. Conclusion

Collective Intelligence Systems represent an important evolution in competition law because they can transform dispersed market information into collective decision-making power.

The central antitrust danger is not simply artificial intelligence itself. It is the possibility that AI, shared data, common algorithms or digital intermediaries may reduce the competitive independence of participating firms.

The major principles emerging from Topkins, Eturas, Samir Agarwal, RealPage, Cornish-Adebiyi, T-Mobile Netherlands, AC-Treuhand and Apple can be summarised as follows:

1. An algorithm does not provide immunity from competition law.
2. A common digital platform can facilitate concerted conduct.
3. Algorithmic pricing alone does not automatically prove an agreement.
4. Competitively sensitive data requires particular caution.
5. Third-party technology providers may become relevant where they facilitate coordination.
6. Hub-and-spoke structures can exist without direct competitor-to-competitor communication.
7. Collective intelligence can create both cartel and dominance concerns.
8. The decisive issue is whether the system preserves or undermines independent competitive decision-making.