AI Ecosystem Co-Adaptation and Tacit Coordination — Detailed Explanation with Case Laws

1. Introduction

AI ecosystem co-adaptation refers to a situation in which competing firms, platforms, developers, suppliers, distributors, or service providers continuously adjust their conduct in response to one another while using interconnected AI systems, common datasets, APIs, benchmarks, cloud infrastructure, recommendation engines, pricing systems, or third-party algorithms.

Tacit coordination is different from an express cartel. Firms may reach similar commercial outcomes without an explicit agreement, for example by repeatedly observing rivals' algorithmic behaviour and adapting their own strategies accordingly.

AI can intensify this problem because algorithms can:

• monitor competitors continuously;
• process large quantities of market information;
• rapidly change prices or terms;
• learn from competitors' actions;
• predict likely competitive responses;
• use common third-party datasets;
• optimise toward similar objectives;
• create network effects and feedback loops;
• make market deviations immediately observable; and
• reduce the need for direct human communication.

The important legal distinction is that parallel or adaptive conduct is not automatically unlawful coordination. Competition law generally requires an agreement, concerted practice, coordinated conduct, or—depending on the legal provision—conduct amounting to collective dominance or another prohibited abuse.

2. Meaning of AI Ecosystem Co-Adaptation

An AI ecosystem may contain several interconnected layers:

Data → Cloud/Compute → Foundation Model → Algorithm → Platform → Businesses → Consumers → Behavioural Data → AI Model

Each participant can adapt to changes elsewhere in the ecosystem.

For example:

Platform A changes its recommendation algorithm → sellers alter prices → Platform B observes those prices → B's AI changes its recommendations → sellers respond again.

This can produce convergent market behaviour without a conventional cartel meeting.

The competition concern becomes stronger when:

1. the market has few significant competitors;
2. firms have access to highly transparent real-time information;
3. algorithms repeatedly observe one another;
4. competitors use the same pricing or optimisation provider;
5. competitively sensitive information is shared;
6. deviations are rapidly detected;
7. algorithms punish deviation;
8. firms repeatedly use identical strategic parameters; or
9. the ecosystem creates structural dependence on one intermediary.

3. Tacit Coordination versus Express Collusion

The fundamental problem is therefore:

When does independent algorithmic adaptation remain legitimate competition, and when does it become evidence of a concerted practice or facilitate unlawful coordination?

4. Legal Framework

A. Agreement or Concerted Practice

Under EU competition law, Article 101 TFEU covers:

• agreements;
• decisions by associations of undertakings; and
• concerted practices.

The central issue is whether competitors have knowingly substituted practical cooperation for the risks of independent competition.

AI does not create an exemption from this principle.

B. U.S. Antitrust Law

Section 1 of the Sherman Act prohibits contracts, combinations and conspiracies restraining trade.

A particularly difficult issue with AI is the agreement requirement.

If competitors independently instruct their algorithms to maximise profits, similar outcomes alone do not necessarily establish a conspiracy.

But if competitors:

• agree to use the same algorithm for coordination;
• provide competitively sensitive information to the algorithm;
• agree on pricing parameters;
• use an intermediary to coordinate conduct; or
• deliberately use software to implement an existing agreement,

traditional antitrust principles can apply.

The U.S. Department of Justice's RealPage proceedings demonstrate this modern application. The DOJ alleged that competing landlords supplied competitively sensitive information to a common pricing system and used algorithmic recommendations that aligned pricing.

5. Six Important Case Laws

Case 1 — T-Mobile Netherlands BV v Raad van bestuur van de Nederlandse Mededingingsautoriteit

Case: C-8/08
Court: Court of Justice of the European Union
Year: 2009

Facts

Representatives of competing mobile operators attended a meeting where commercially sensitive information concerning market conduct was discussed.

The issue was whether a single meeting could constitute a concerted practice.

Principle

The Court held that a single meeting can, depending on its content and circumstances, be sufficient to establish a concerted practice.

The Court also recognised the importance of the causal relationship between the concerted conduct and subsequent market behaviour.

AI relevance

This case is important because AI coordination does not necessarily require continuous communications.

For example:

One coordinated technical meeting → agreement on algorithmic parameters → continuous algorithmic implementation.

The subsequent absence of human communications would not necessarily eliminate the legal significance of the initial coordination.

Key lesson

The absence of continuing human communication does not necessarily mean that subsequent algorithmic conduct is independent.

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

Case: C-74/14
Court: CJEU
Year: 2016

Facts

Several travel agencies used a common computerised booking system.

The system administrator introduced an automatic restriction on the discounts that agencies could offer online. The administrator also sent a communication concerning the restriction.

The question was whether the agencies' subsequent conduct could constitute a concerted practice.

Principle

The CJEU considered whether knowledge of the system administrator's conduct, combined with continued participation in the system, could support an inference of concerted conduct.

The case specifically concerned a common computerised booking system and automatic restriction of discounts, making it exceptionally relevant to algorithmic competition analysis.

AI relevance

The modern equivalent could involve:

• a common AI pricing platform;
• common recommendation software;
• a shared optimisation engine;
• automatically imposed pricing floors; or
• common discount restrictions.

Key lesson

A common technological infrastructure can become legally significant when competitors know that it is being used to restrict competitive parameters.

Case 3 — United States v David Topkins

Court: U.S. District Court / Department of Justice enforcement
Year: 2015

Facts

David Topkins and other online sellers of posters allegedly agreed to coordinate prices on Amazon Marketplace.

They did not merely communicate about prices. They agreed to use pricing algorithms to implement their coordination.

Topkins pleaded guilty to price-fixing charges. The case is widely regarded as an early U.S. example of algorithm-assisted price coordination.

AI relevance

The distinction here is crucial:

Algorithm itself ≠ cartel

but:

Agreement to use an algorithm to implement price fixing = conventional cartel conduct implemented through technology.

Key lesson

Competition law follows the substantive conduct, not the technological mechanism.

An algorithm cannot legitimise an otherwise unlawful price-fixing arrangement.

Case 4 — Airtours v Commission

Case: T-342/99
Court: Court of First Instance of the European Communities
Year: 2002

Facts

The proposed Airtours/First Choice merger concerned the UK package-holiday market.

The European Commission considered that the merger could create a collective dominant position because the remaining large firms could coordinate their behaviour without an express cartel.

The Commission's decision was annulled because the Court found that the Commission had not sufficiently established the necessary conditions for collective dominance.

Importance for tacit coordination

The case developed the classic framework concerning tacit coordination.

Important considerations include:

1. whether competitors can observe each other's behaviour;
2. whether coordination can be sustained;
3. whether deviations can be detected;
4. whether retaliation is possible; and
5. whether external competition would destabilise coordination.

AI relevance

AI can dramatically increase these capabilities.

A conventional market might allow competitors to observe prices once a day.

An AI ecosystem may allow:

continuous monitoring → instant detection → automated response → repeated adjustment.

Thus, AI can potentially make some traditional conditions for tacit coordination easier to satisfy.

Key lesson

Market transparency, monitoring and credible responses are central to analysing tacit coordination.

Case 5 — Impala v Commission / Sony-BMG

Case: T-464/04
Court: Court of First Instance of the European Communities
Year: 2006

Facts

The case concerned the proposed Sony-BMG merger and the possibility of collective dominance in the recorded-music market.

The Court examined whether market characteristics could facilitate tacit coordination.

The judgment recognised that evidence such as sustained price alignment, stable market shares and other structural indicators could be relevant to determining whether a market permits tacit coordination.

AI relevance

AI can make the types of market transparency examined in Impala significantly more sophisticated.

For example:

• real-time price monitoring;
• demand forecasting;
• competitor inventory prediction;
• automated capacity adjustment;
• customer segmentation; and
• automated retaliation.

Key lesson

Persistent alignment should be analysed together with market structure and alternative explanations.

AI-generated parallel behaviour alone should not automatically be treated as proof of unlawful coordination.

Case 6 — United States v RealPage, Inc.

Court: U.S. District Court for the Middle District of North Carolina
Proceedings: 2024–2026

This is one of the most directly relevant contemporary examples.

Facts

The DOJ alleged that RealPage's revenue-management software used competitively sensitive information supplied by landlords to generate rental pricing recommendations.

The government alleged that landlords used the system in a manner that reduced independent price competition. The case involves claims under Sections 1 and 2 of the Sherman Act.

The DOJ subsequently obtained proposed settlements involving RealPage and several landlord defendants addressing information sharing and algorithmic coordination.

AI relevance

The case illustrates a particularly important model:

Competitor data

↓

Common algorithm

↓

Algorithmic recommendation

↓

Competitor adoption

↓

Reduced independent decision-making

↓

Potentially coordinated market outcome

Key lesson

A third-party AI or algorithm provider may become an important part of the competition-law analysis where competitors provide it with competitively sensitive information and use its outputs to make coordinated commercial decisions.

6. Additional Case: Wood Pulp

A. Ahlström Osakeyhtiö and Others v Commission

Joined Cases: C-89/85 and others
Court: CJEU
Year: 1993

This line of authority is important for distinguishing parallel conduct from an unlawful concerted practice.

The fact that competitors behave similarly does not automatically prove that they coordinated their behaviour.

AI relevance

This becomes extremely important in machine-learning markets.

Suppose competing AI systems independently learn that:

• demand is rising;
• a competitor has increased prices;
• capacity is constrained; and
• raising prices maximises expected profit.

All systems may therefore increase prices.

That is not automatically a cartel.

Authorities must examine whether there is evidence of communication, information exchange, common coordination mechanisms, or other factors demonstrating substitution of cooperation for independent competition.

7. How AI Can Facilitate Tacit Coordination

7.1 Continuous Monitoring

Traditional firms may monitor competitors periodically.

AI systems can monitor:

• prices;
• inventories;
• advertisements;
• promotions;
• product launches;
• customer behaviour;
• search rankings;
• delivery times;
• capacity; and
• competitor algorithmic responses.

This makes deviations easier to detect.

7.2 Algorithmic Retaliation

An AI system can potentially detect:

Competitor lowers price → automatically lower price.

Or:

Competitor increases capacity → automatically change capacity.

If repeated interaction produces stable market outcomes, regulators may examine whether the system creates conditions conducive to coordination.

8. Common Algorithm Providers

One of the most important risks arises when several competitors use the same third-party AI provider.

Example:

Competitor A

↘

Common AI Pricing Platform

↗

Competitor B

If the provider receives sensitive information from both competitors, the platform may become a central information hub.

Potential risks include:

• exchange of non-public pricing data;
• common pricing recommendations;
• common demand forecasts;
• common discount parameters;
• common optimisation rules;
• common inventory information.

The RealPage litigation illustrates why authorities are increasingly examining this structure.

9. Feedback-Loop Problem

AI ecosystems can create a self-reinforcing feedback loop:

Competitor prices

↓

AI observes prices

↓

AI predicts competitor behaviour

↓

AI recommends price

↓

Firm changes price

↓

Competitor AI observes change

↓

Competitor AI responds

↓

New data enters the system

The cycle repeats.

This may produce highly stable prices without traditional cartel meetings.

The legal question is therefore not simply:

"Did the algorithm communicate with another algorithm?"

Instead:

What human or corporate decisions created, configured, supplied, supervised, or knowingly adopted the algorithmic coordination mechanism?

10. Common Data Pools and Competition Risks

AI systems require enormous datasets.

If competing companies contribute their own sensitive information to a common AI system, the following information may become competitively problematic:

• current prices;
• future prices;
• costs;
• capacity;
• inventories;
• margins;
• customer-specific information;
• production plans;
• bidding strategies;
• promotional plans.

The competitive concern increases when the information is:

current + granular + non-public + competitor-specific + strategically useful.

11. AI Ecosystem Co-Adaptation in Digital Markets

The phenomenon can arise beyond pricing.

Search

AI search systems may independently adjust rankings in response to competing search engines.

Advertising

Advertising algorithms may learn similar bidding strategies.

E-commerce

Recommendation systems may react to competing sellers' inventory and prices.

Ride-hailing

Algorithms may respond to competing platforms' prices and driver availability.

Cloud computing

AI systems may optimise capacity and pricing based upon competitors' capacity.

App stores

Platforms may adjust commissions, rankings and access conditions based upon developers' behaviour.

Generative AI

Foundation-model providers may respond to competitors':

• model releases;
• API prices;
• context windows;
• safety restrictions;
• benchmark performance; and
• licensing terms.

These activities are not automatically unlawful. The relevant question is whether competition is being replaced by coordination.

12. Tacit Coordination Through Common AI Objectives

Suppose competing firms independently use AI systems instructed to:

"Maximise long-term profit."

Their algorithms might discover that aggressive price competition is less profitable than maintaining relatively high prices.

The systems could therefore converge toward similar prices.

This presents a difficult legal question.

Scenario A — Independent adaptation

Each company:

• independently develops its AI;
• uses its own data;
• does not communicate with competitors;
• independently chooses parameters.

Parallel prices alone may be insufficient to establish an agreement.

Scenario B — Coordinated implementation

Companies:

• exchange sensitive data;
• use a common algorithm;
• agree on pricing parameters;
• knowingly adopt coordination mechanisms.

The legal risk becomes substantially greater.

Scenario C — Explicit human agreement

Companies agree:

"We will use the AI system to maintain prices above competitive levels."

The algorithm is merely the implementation mechanism.

That resembles conventional cartel conduct, as illustrated by Topkins.

13. Collective Dominance and AI Ecosystems

Tacit coordination is also relevant to collective dominance.

A small number of firms may collectively possess market power where market characteristics enable them to behave in a coordinated manner.

AI can potentially strengthen:

Market transparency

Everyone can observe competitors almost immediately.

Detection

Deviation can be identified automatically.

Retaliation

Algorithms can respond instantly.

Stability

Machine-learning systems can repeatedly optimise around competitors' expected reactions.

These factors correspond closely to the economic conditions examined in Airtours and related collective-dominance jurisprudence.

14. Important Distinction: Parallelism Is Not Automatically Collusion

This is perhaps the most important principle.

Consider:

• Firm A raises its price.
• Firm B observes it.
• Firm B independently raises its price.
• Firm C follows.

That conduct could arise through rational unilateral adaptation.

The existence of similar AI outputs does not itself prove:

• an agreement;
• communication;
• concerted practice;
• collective dominance; or
• unlawful collusion.

Authorities generally need to consider the totality of circumstances.

Relevant evidence can include:

1. communications;
2. common algorithms;
3. data sharing;
4. algorithm configuration;
5. internal documents;
6. pricing instructions;
7. adoption of competitors' strategic information;
8. monitoring systems;
9. retaliation mechanisms;
10. unexplained persistent coordination.

15. Competition Risks Specific to AI Ecosystems

A. Information Exchange

Competitors may indirectly exchange sensitive information through an AI intermediary.

B. Algorithmic Hub-and-Spoke Coordination

A common technology provider can potentially become the hub connecting otherwise competing firms.

C. Reduced Independent Decision-Making

Businesses may simply accept algorithmic recommendations rather than independently determining prices or commercial strategies.

D. Increased Market Transparency

AI may make competitor behaviour observable in real time.

E. Automated Retaliation

Algorithms may automatically respond to deviations.

F. Common Model Risk

Competitors using identical models may converge toward identical commercial strategies.

G. Data Concentration

One AI provider may obtain extensive information from multiple competitors.

H. Network Effects

The larger the dataset, the better the model may become, potentially reinforcing the position of a central AI intermediary.

16. Evidence Authorities May Examine

In an AI-coordination investigation, traditional cartel evidence may be supplemented by technical evidence.

Corporate evidence

• emails;
• board minutes;
• contracts;
• pricing policies;
• internal presentations.

Algorithmic evidence

• source code;
• model architecture;
• prompts;
• system instructions;
• reward functions;
• optimisation objectives;
• training datasets;
• model logs;
• API calls.

Commercial evidence

• price histories;
• discount patterns;
• capacity decisions;
• inventory movements;
• margins.

Technical evidence

• data flows;
• model inputs;
• model outputs;
• version histories;
• audit logs;
• parameter changes.

The RealPage proceedings demonstrate the increasing importance of technical analysis in antitrust enforcement involving algorithmic pricing. The DOJ stated that its investigators used data-science expertise to examine how the algorithms used landlords' sensitive information.

17. Compliance Measures for AI Ecosystems

Companies using AI in competitive markets should consider:

1. Independent-data architecture

Avoid unnecessary ingestion of competitors' sensitive information.

2. Competitor-data restrictions

Establish clear rules governing:

• prices;
• costs;
• capacity;
• future strategy;
• customer information.

3. Algorithm governance

Maintain records of:

• who designed the algorithm;
• what objective it optimises;
• what data it receives;
• what constraints it uses.

4. Human override

Commercial decisions should not necessarily be automatically implemented merely because an algorithm recommends them.

5. Third-party AI due diligence

Companies should understand whether their AI provider:

• aggregates competitor data;
• uses it for model training;
• creates cross-customer recommendations;
• permits data isolation.

6. Auditability

Maintain sufficient logs to reconstruct:

input → model processing → recommendation → human decision → market action.

18. Legal Test for AI Co-Adaptation

A useful analytical framework is:

Step 1 — Identify the relevant market

Determine:

• product market;
• geographic market;
• competitors;
• market concentration.

Step 2 — Identify the AI relationship

Ask whether firms:

• independently developed AI;
• use the same provider;
• use shared infrastructure;
• share datasets.

Step 3 — Examine information flows

Determine whether competitively sensitive information moves between competitors.

Step 4 — Examine human involvement

Ask:

Did managers know how the system worked?

Did they intentionally configure it to facilitate coordination?

Step 5 — Examine algorithmic behaviour

Analyse:

• price convergence;
• output convergence;
• reaction speed;
• deviation detection;
• retaliation.

Step 6 — Identify alternative explanations

Consider:

• common demand shocks;
• common costs;
• market-wide inflation;
• capacity constraints;
• legitimate optimisation.

Step 7 — Determine the legal theory

Potential theories include:

• agreement;
• concerted practice;
• information exchange;
• cartel;
• hub-and-spoke coordination;
• collective dominance;
• abuse of dominance;
• merger-related coordinated effects.

19. Comparative Case-Law Lessons

20. Key Legal Principle

The central rule can be expressed as:

AI may change the mechanism of coordination, but it does not change the underlying competition-law question.

The decisive inquiry is not whether:

"The machines coordinated."

Rather, authorities should investigate:

Who designed the system, who supplied the data, what information was exchanged, what objectives were imposed, what decisions were knowingly delegated, and whether the resulting conduct replaced independent competitive decision-making with coordinated behaviour?

The distinction between independent algorithmic adaptation and algorithm-facilitated concerted conduct is therefore fundamental.

21. Conclusion

AI ecosystem co-adaptation creates a new competition-law problem because interconnected algorithms can make markets:

• more transparent;
• faster-moving;
• highly responsive;
• data-intensive;
• increasingly dependent on common infrastructure; and
• potentially more conducive to stable coordination.

However, similar AI outputs do not by themselves establish unlawful collusion. The legal analysis must distinguish ordinary interdependence and rational adaptation from conduct in which competitors knowingly coordinate, exchange competitively sensitive information, use a common mechanism to restrict competition, or otherwise substitute cooperation for independent market decision-making.

The most relevant authorities—particularly T-Mobile Netherlands, Eturas, Topkins, Airtours, Impala and RealPage—show the evolution from traditional human coordination toward increasingly technology-mediated forms of coordination. The RealPage litigation is especially significant because it demonstrates that competition authorities are applying conventional antitrust principles to modern algorithmic pricing systems rather than treating AI as a separate legal category.

Thus, the emerging principle is:

Independent AI adaptation → generally requires careful contextual analysis.

AI used to implement an agreement → conventional antitrust risk.

Shared sensitive data + common algorithm + coordinated commercial conduct → substantial competition-law concern.

AI-enabled structural transparency + monitoring + retaliation → potentially relevant to tacit-coordination/collective-dominance analysis.