Competition Law and Competition Governance of Intelligence Marketplaces

1. Introduction

Intelligence marketplaces are emerging markets in which artificial intelligence capabilities are bought, sold, licensed, accessed, or allocated through platforms. The relevant “intelligence” may include:

• foundation and generative AI models;
• AI agents and autonomous systems;
• model APIs and inference services;
• training and inference compute;
• datasets and proprietary data;
• AI-generated content and synthetic data;
• model fine-tuning and evaluation services;
• AI plugins, tools and agentic applications;
• AI-as-a-Service and cloud-based intelligence;
• marketplaces matching AI developers with businesses or consumers.

These marketplaces are economically unusual because competition may occur simultaneously at several layers: compute → data → foundation models → APIs → applications/agents → users.

Competition law therefore has to examine not merely whether an AI firm is large, but whether it can use control over one layer to foreclose competitors at another layer, discriminate between marketplace participants, restrict interoperability, acquire emerging rivals, or use data and algorithms to entrench market power.

The UK CMA's work on AI foundation models identifies access to data, compute, expertise and funding, together with switching costs and vertically integrated partnerships, as significant competition considerations. The FTC has similarly examined large cloud-AI partnerships for possible effects on access to compute and talent, switching costs and access to commercially sensitive information.

2. Meaning of an Intelligence Marketplace

An intelligence marketplace can be represented as:

Inputs

Data + Compute + Chips + Talent + Capital

↓

Intelligence Production

Foundation Model / AI Model / Agent

↓

Distribution Layer

API + Cloud + App Store + AI Marketplace

↓

Applications

Search + Coding + Healthcare + Finance + Robotics + Commerce

↓

Users

Consumers + Enterprises + Governments

The competition problem is that a single undertaking may control several of these layers simultaneously.

For example, a cloud provider may:

1. supply computing infrastructure;
2. invest in an AI developer;
3. operate its own foundation model;
4. operate an AI marketplace;
5. distribute competing AI products;
6. possess information concerning rival AI developers.

This creates opportunities for vertical leverage.

3. Relevant Competition-Law Framework

A. Market Definition

Traditional competition law may need to identify several interrelated markets.

Possible relevant markets

1. AI compute;
2. cloud infrastructure;
3. foundation models;
4. model APIs;
5. AI-agent marketplaces;
6. AI application distribution;
7. AI data;
8. AI evaluation and verification services;
9. enterprise AI services;
10. consumer AI assistants.

The relevant question is not necessarily whether “AI” constitutes one market.

Instead, authorities may examine whether a particular service constitutes a distinct relevant product market and whether adjacent markets are sufficiently connected to permit leverage.

4. Multi-Sided Market Structure

Intelligence marketplaces are frequently multi-sided platforms.

An AI marketplace can simultaneously connect:

• model developers;
• application developers;
• enterprises;
• consumers;
• data providers;
• compute providers;
• advertisers;
• AI-agent developers.

This makes network effects especially important.

More developers may attract more users.

More users generate more interactions and potentially more data.

More data can improve models.

Better models attract more developers.

This creates a data–model–user feedback loop.

The Supreme Court's decision in Ohio v. American Express Co., 585 U.S. 529 (2018) is important by analogy because it recognized the distinctive characteristics of two-sided transaction platforms and required competitive effects to be considered across the platform rather than mechanically examining only one side.

For intelligence marketplaces, this means that an authority may need to examine both:

AI developers ↔ marketplace ↔ users

rather than looking exclusively at developer fees or consumer prices.

5. Dominance and Market Power

Market power in intelligence marketplaces may arise from:

1. Compute concentration

Large-scale AI training may require enormous computational resources.

2. Data advantages

Incumbents may possess proprietary datasets unavailable to competitors.

3. Network effects

A marketplace with more users may attract more AI developers, while more developers increase user value.

4. Switching costs

Enterprise customers may become dependent on:

• proprietary APIs;
• model-specific tooling;
• cloud infrastructure;
• embeddings;
• databases;
• fine-tuning systems.

5. Ecosystem integration

A firm controlling:

operating system + cloud + search + browser + AI assistant + app marketplace

may have opportunities to extend power from one market into another.

6. Reputation and trust

AI systems may benefit from accumulated reputation, enterprise certification and reliability records.

6. Major Competition Concerns

A. Self-Preferencing

An intelligence marketplace operator may rank its own AI model above competing models.

For example:

Marketplace → search results → own model automatically appears first.

The legal concern is comparable to preferential treatment of an undertaking's own downstream service.

The Google Shopping litigation is particularly relevant. In Google and Alphabet v Commission, Case C-48/22 P (2024), the Court of Justice upheld the finding concerning Google's preferential treatment of its own comparison-shopping service in search results.

Application to intelligence marketplaces

Potentially problematic conduct could include:

• ranking an affiliated model first;
• preferential API access;
• lower fees for affiliated applications;
• privileged access to marketplace data;
• superior computing allocation;
• preferential recommendation by an AI assistant.

The important legal question would remain whether the conduct constitutes exclusionary abuse rather than simply preferential competition on the merits.

7. Tying and Bundling

An intelligence marketplace may tie:

AI assistant + operating system

or:

AI model + cloud services

or:

AI API + proprietary database

or:

AI agent + payment system.

The classic digital-platform precedent is Google Android, Case T-604/18 (2022).

The General Court largely upheld findings concerning Google's restrictions involving Android, Google Search, Chrome, Play Store and device manufacturers, while reducing the fine to €4.125 billion.

Relevance to AI

An analogous concern could arise where an undertaking with substantial power in one layer makes access to an essential or highly valuable service conditional upon adoption of its AI product.

8. Exclusive Dealing

Intelligence marketplaces may use:

• exclusive cloud arrangements;
• exclusive AI distribution;
• default AI-assistant agreements;
• exclusive model hosting;
• restrictions on competing models.

The concern becomes particularly serious where the platform is an important gateway.

The Google Search litigation demonstrates how distribution agreements can become central to monopolization analysis. In the United States, remedies ordered in 2025 included restrictions on certain exclusive distribution arrangements and provisions requiring specified access to search data and syndication services.

AI analogy

Comparable concerns could arise if:

a dominant device/platform → makes its AI assistant the default → restricts competing assistants.

9. Data Access and Data Advantage

Data may become a strategic competitive input.

A dominant intelligence marketplace may obtain:

• user prompts;
• search histories;
• transaction information;
• developer performance data;
• model usage information;
• enterprise data;
• feedback and evaluation data.

If the platform simultaneously competes with marketplace participants, it may possess information unavailable to those competitors.

The Amazon Marketplace commitments are highly relevant. The European Commission's investigation concerned Amazon's use of non-public seller data, Buy Box selection and Prime-related practices. Amazon ultimately undertook commitments concerning seller data, Buy Box presentation and equal treatment.

AI application

An AI marketplace operator might obtain confidential information about:

which AI applications are growing → which customers they serve → their pricing → conversion rates → model performance.

Using that information to improve an affiliated competing service could create significant competition concerns.

10. Interoperability

Interoperability may be critical for intelligence marketplaces.

Potential restrictions include:

• preventing models from operating on rival clouds;
• restricting agent-to-agent communication;
• preventing data portability;
• blocking competing AI assistants from operating through an operating system;
• refusing API interoperability.

The EU's 2026 DMA work illustrates the increasing regulatory importance of AI interoperability. The European Commission issued binding specification measures concerning access by competing AI services to certain Android functionalities.

Thus, interoperability can become a competition remedy rather than merely a technical design choice.

11. Switching Costs and Lock-In

AI ecosystems can generate unusually high switching costs.

An enterprise may build:

databases → prompts → fine-tuning → workflows → agents → employee training

around one provider.

Switching may therefore require:

• retraining;
• data migration;
• API modification;
• security re-certification;
• workflow reconstruction;
• model re-evaluation.

The FTC's investigation into major cloud-AI partnerships specifically identified potential increases in contractual and technical switching costs as a competition issue.

Competition governance therefore needs to consider contestability over time, not merely current market shares.

12. Vertical Integration

A particularly important structural issue is:

Cloud + Model + Marketplace + Application

For example, a company may simultaneously be:

• cloud infrastructure provider;
• AI-model developer;
• AI marketplace operator;
• enterprise application provider.

It could theoretically discriminate against independent models through:

• compute pricing;
• latency;
• API access;
• search ranking;
• data access;
• technical integration;
• contractual restrictions.

The FTC's AI partnerships study specifically examined the relationships between major cloud providers and leading AI developers, including Microsoft–OpenAI, Amazon–Anthropic and Google–Anthropic.

13. AI Partnerships and Minority Investments

Competition authorities increasingly need to scrutinize transactions that do not resemble traditional acquisitions.

Examples include:

• minority investments;
• revenue-sharing agreements;
• exclusive cloud commitments;
• long-term compute agreements;
• preferred distribution arrangements;
• licensing agreements;
• board or governance rights.

The concern is that a large technology company may obtain substantial strategic influence without formally acquiring the AI company.

The FTC's 2025 report specifically highlighted equity and revenue-sharing rights, consultation/control rights and exclusivity provisions in major AI partnerships.

14. Merger Control

Traditional turnover-based merger thresholds can miss acquisitions of emerging AI competitors because startups may have:

• low current revenues;
• valuable technology;
• valuable data;
• highly skilled personnel;
• rapid growth potential.

Competition governance should therefore examine:

Killer-acquisition risks

A dominant AI platform acquires an emerging model or agent that could become a competitive constraint.

Capability acquisition

Acquiring:

• model technology;
• engineers;
• patents;
• datasets;
• agents.

Ecosystem acquisition

An AI platform purchases complementary technologies to make its ecosystem harder to challenge.

The FTC's Meta/Within litigation illustrates the concern about acquisitions in emerging technology markets where the authority alleged that acquisition of a developing VR fitness application could reduce future innovation and competition.

The same theory can be relevant to AI acquisitions, although each transaction requires its own evidence.

15. Essential-Facility and Access Questions

Some intelligence marketplaces may control particularly important inputs:

• compute;
• specialized chips;
• proprietary datasets;
• model interfaces;
• distribution channels;
• AI-agent operating environments.

A refusal to supply is not automatically unlawful.

Competition law generally requires a demanding analysis of:

1. dominance;
2. indispensability;
3. absence of reasonable alternatives;
4. exclusionary effects;
5. justification;
6. feasibility of access;
7. effects on incentives to invest.

Thus, the fact that an AI company controls an important resource does not by itself establish an obligation to provide access.

16. Algorithmic Discrimination

An AI marketplace may employ algorithms to determine:

• ranking;
• recommendations;
• model visibility;
• API allocation;
• prices;
• commissions;
• access.

This creates the possibility of algorithmic discrimination.

For example:

affiliated AI model → preferential ranking → more users → more feedback → better model → greater ranking advantage.

This can generate a self-reinforcing competitive loop.

Competition authorities therefore need to examine not only the output of algorithms but also:

• training data;
• ranking criteria;
• optimization objectives;
• exclusion rules;
• affiliate treatment;
• access to marketplace data.

17. Algorithmic Collusion

Intelligence marketplaces could also facilitate coordination.

AI systems may automatically observe:

• competitors' prices;
• capacity;
• demand;
• inventory;
• commissions;
• model usage.

Algorithms could potentially facilitate parallel pricing without conventional human communication.

However, mere parallel conduct is not automatically an antitrust agreement.

Authorities would need to distinguish:

independent algorithmic adaptation

from:

unlawful concerted coordination.

This distinction will become increasingly important as autonomous AI systems participate in commercial decision-making.

18. Case Laws

1. Google Shopping — Google and Alphabet v Commission, C-48/22 P (2024)

Principle

Google's preferential positioning of its own comparison-shopping service in general search results was found to constitute abusive conduct under Article 102 TFEU.

Intelligence-marketplace relevance

The case provides a strong framework for analysing:

• AI marketplace self-preferencing;
• preferential model ranking;
• AI-agent recommendation bias;
• affiliated application promotion.

The key lesson is that an intermediary may face competition-law scrutiny when it uses a powerful gateway to favor its own downstream service.

2. Google Android — Google and Alphabet v Commission, T-604/18 (2022)

Principle

The General Court largely upheld the Commission's findings concerning Google's restrictions involving Android, including tying/bundling, exclusivity-related arrangements and anti-fragmentation obligations.

Intelligence-marketplace relevance

It provides guidance for analysing:

• AI/cloud bundling;
• AI assistant defaults;
• model-platform integration;
• restrictions on competing AI services;
• ecosystem leveraging.

3. Ohio v American Express Co., 585 U.S. 529 (2018)

Principle

The Supreme Court treated American Express as a two-sided transaction platform and required the competitive analysis to consider both sides of the platform.

Intelligence-marketplace relevance

An AI marketplace may similarly connect:

developers ↔ users

or:

models ↔ applications ↔ enterprises.

Competition effects therefore may need to be assessed across the interconnected ecosystem.

4. FTC v Facebook/Meta, No. 20-cv-3590

Principle

The FTC alleged that Facebook maintained monopoly power through acquisitions of Instagram and WhatsApp and restrictions involving software developers and API access. The litigation remains active following the FTC's appeal from a 2025 district-court ruling.

Intelligence-marketplace relevance

The case illustrates competition concerns involving:

• nascent competitors;
• acquisitions;
• APIs;
• platform access;
• ecosystem entrenchment.

These are directly relevant to AI marketplaces where emerging AI agents or models may become future competitive constraints.

5. United States v Google LLC — Search Monopolization

Principle

The U.S. litigation concerned Google's alleged maintenance of monopolies through exclusionary distribution arrangements.

The 2025 remedies included restrictions on certain exclusive arrangements and requirements concerning access to specified search data and syndication services.

Intelligence-marketplace relevance

It demonstrates the potential importance of:

• defaults;
• distribution;
• data access;
• interoperability;
• exclusionary contracts.

The same concepts can arise when AI assistants become default interfaces.

6. Amazon Marketplace — European Commission

Principle

The Commission's investigation addressed Amazon's use of non-public seller data, Buy Box practices and aspects of Prime and logistics.

Amazon's commitments included restrictions on using non-public seller data, changes to Buy Box presentation and measures supporting greater neutrality and choice for sellers.

Intelligence-marketplace relevance

This is particularly important for AI marketplaces because the platform may simultaneously:

operate the marketplace + observe marketplace participants + compete with them.

That creates a potential platform-as-referee-and-player problem.

7. Epic Games v Google

The Epic litigation concerning Google Play provides another useful digital-platform precedent concerning app distribution, payments, platform access and two-sided markets.

The litigation illustrates how control over a digital distribution gateway can generate competition-law disputes concerning:

• platform fees;
• alternative distribution;
• payment systems;
• developer access;
• interoperability.

The Ninth Circuit litigation has continued into 2025–26, making it a particularly relevant contemporary platform precedent.

19. Competition Governance Model for Intelligence Marketplaces

A sophisticated regulatory framework can be structured around eight pillars.

Pillar 1 — Access

Ensure reasonable competitive access to:

• compute;
• data;
• APIs;
• technical interfaces;
• distribution.

Pillar 2 — Interoperability

Promote:

• API interoperability;
• data portability;
• model portability;
• agent interoperability.

Pillar 3 — Neutrality

Monitor:

• self-preferencing;
• discriminatory ranking;
• preferential API treatment;
• affiliated-model advantages.

Pillar 4 — Transparency

Require appropriate transparency concerning:

• ranking criteria;
• access conditions;
• commissions;
• marketplace rules;
• model selection.

Transparency should not require disclosure of legitimate trade secrets.

Pillar 5 — Contestability

Monitor:

• switching costs;
• exclusivity;
• defaults;
• technical lock-in;
• contractual restrictions.

Pillar 6 — Merger Control

Examine acquisitions involving:

• emerging AI competitors;
• valuable datasets;
• specialized models;
• AI talent;
• strategically important agents.

Pillar 7 — Algorithmic Governance

Audit for:

• discriminatory ranking;
• exclusionary algorithms;
• coordinated pricing;
• automated foreclosure.

Pillar 8 — Remedy Design

Possible remedies include:

• non-discrimination;
• interoperability;
• data access;
• portability;
• API access;
• structural separation;
• behavioural commitments;
• monitoring trustees;
• restrictions on exclusivity.

20. India-Specific Perspective

For India, the principal statutory framework is the Competition Act, 2002, particularly:

• Section 3 — anti-competitive agreements;
• Section 4 — abuse of dominant position;
• Section 5 — combinations;
• Section 6 — regulation of combinations;
• Sections 19–27 — investigation and remedies.

An intelligence marketplace could potentially raise Section 4 concerns where a dominant digital undertaking engages in:

• discriminatory access;
• unfair conditions;
• tying;
• leveraging;
• denial of market access;
• exclusionary contracts;
• self-preferencing.

Section 3 could become relevant where competing AI providers coordinate:

• prices;
• access conditions;
• commissions;
• model restrictions;
• output or development.

Combination control may become relevant where major technology companies acquire AI developers or obtain strategically significant control through investment and partnership arrangements.

21. Special Problem: Intelligence as Both Product and Competitor

The most distinctive competition problem is that intelligence itself can become the intermediary.

An AI agent may decide:

which seller → which product → which service → which model → which payment method

to present to a consumer.

Therefore, the AI system can become a gatekeeper between competing businesses and customers.

If the agent is controlled by a vertically integrated company, the company may possess incentives to direct users toward its own:

• products;
• models;
• applications;
• payment systems;
• advertising services.

This transforms AI competition from a conventional product-market issue into a gateway governance problem.

22. Competition Governance of Autonomous Intelligence Marketplaces

Future marketplaces may contain autonomous agents capable of:

• negotiating contracts;
• selecting suppliers;
• purchasing compute;
• selecting models;
• changing prices;
• reallocating resources;
• negotiating API access.

This raises a new question:

Who is legally responsible when an autonomous intelligence system makes an exclusionary commercial decision?

Potentially relevant actors include:

1. marketplace operator;
2. AI model provider;
3. deploying enterprise;
4. agent developer;
5. human decision-maker;
6. algorithmic intermediary.

Competition governance therefore needs clear rules on attribution, auditability and accountability.

23. Key Competition Risks — Summary Table

24. Six Core Legal Principles

The emerging law of intelligence marketplaces can therefore be summarized through six principles:

1. Market power can exist at multiple layers

Control over compute, data, models or distribution may produce leverage into adjacent markets.

2. Platform neutrality becomes increasingly important

A marketplace that competes with its own participants creates a potential conflict of interest.

3. Data can be a competitive input

Exclusive access to commercially significant data may reinforce market power.

4. Interoperability can determine contestability

Closed technical ecosystems can increase switching costs and entry barriers.

5. AI partnerships require competition scrutiny

Investment and contractual relationships can produce effects similar to vertical integration even without a conventional merger. The FTC's AI partnerships study specifically identified these concerns.

6. Competition governance must be dynamic

The CMA's AI work emphasizes access, diversity, choice and contestability because AI markets can develop rapidly and network effects can become entrenched.

25. Conclusion

Intelligence marketplaces represent a convergence of platform economics, AI, data, cloud infrastructure and automated decision-making. Their competition problems are therefore unlikely to be confined to conventional price-based analysis.

The principal legal challenges will involve:

market definition + multi-sided platforms + data + compute + interoperability + self-preferencing + vertical integration + AI partnerships + exclusionary contracts + algorithmic decision-making.

The most important precedents—Google Shopping, Google Android, Ohio v American Express, FTC v Facebook/Meta, United States v Google, Amazon Marketplace and Epic Games v Google—provide different pieces of the analytical framework.

The emerging regulatory approach is increasingly focused on keeping intelligence markets contestable, interoperable and open, while preserving legitimate incentives to innovate and invest. The joint 2024 U.S.–EU–UK competition statement similarly identified fair dealing, interoperability and choice as central principles for competitive AI markets.

Thus, competition governance of intelligence marketplaces is ultimately concerned with preventing a transition from:

“many firms competing to provide intelligence”

into:

“a small number of vertically integrated gateways controlling who can develop, distribute, access and consume intelligence.”