Competition Law and Competition Concerns in Machine Creativity Markets

1. Introduction

Machine creativity markets refer to markets in which artificial intelligence systems generate, assist in generating, or materially transform creative outputs such as text, music, images, video, software code, advertising content, designs, games and other intellectual or cultural products.

These markets differ from conventional digital markets because competition can occur at several interconnected levels:

1. Creative-input markets — training data, copyrighted works, datasets and human feedback.
2. Compute markets — GPUs, accelerators, cloud infrastructure and model-training capacity.
3. Foundation-model markets — large language, image, audio, video and multimodal models.
4. AI-tool markets — creative assistants, image generators, coding tools and editing platforms.
5. Distribution markets — app stores, search engines, social networks, cloud platforms and marketplaces.
6. Output markets — AI-generated books, music, films, advertisements, software, games and other creative goods.
7. Feedback markets — user interactions, prompts, ratings, edits and behavioural data used to improve models.

The FTC has specifically identified data, computing resources, specialized chips, talent and other AI building blocks as potential competitive bottlenecks.

Accordingly, competition law in machine creativity is not merely concerned with whether an AI company charges a high price. It must also examine innovation, access to inputs, interoperability, data accumulation, foreclosure of rival models, self-preferencing, tying, exclusivity and control over distribution.

2. Meaning of Machine Creativity Markets

Machine creativity exists where an AI system performs functions traditionally associated with human creative activity.

Examples include:

• AI-generated artwork;
• AI-written articles and books;
• AI-generated music;
• AI video generation;
• AI-assisted filmmaking;
• AI advertising and marketing;
• AI-generated game assets;
• AI software programming;
• AI architectural and industrial design;
• synthetic voices and digital performers;
• automated journalism;
• generative design;
• AI-assisted scientific and technical writing.

A useful economic model is:

Data → Compute → Foundation Model → Creative Application → Distribution → User Feedback → Improved Model

Competition concerns may arise at every stage.

3. Relevant Competition-Law Framework

A. Abuse of Dominant Position

A dominant AI firm may potentially violate competition law where it uses its position in one market to exclude competitors in another.

Typical theories include:

• refusal to supply;
• discriminatory access;
• tying and bundling;
• exclusive dealing;
• self-preferencing;
• predatory pricing;
• margin squeeze;
• leveraging;
• discriminatory interoperability;
• exploitation of commercially valuable data.

Under Indian law, these issues principally arise under Section 4 of the Competition Act, 2002.

Under EU law, the corresponding framework principally involves Article 102 TFEU.

Under US law, comparable conduct may fall under Sherman Act §§1–2 and FTC Act §5, depending upon the conduct and market circumstances.

4. Competition Concerns in Machine Creativity Markets

4.1 Concentration of Training Data

High-quality creative AI requires enormous datasets.

A company controlling:

• search queries,
• books,
• photographs,
• videos,
• music,
• social-media interactions,
• consumer behaviour,
• professional databases,

may possess an advantage over new entrants.

The competitive problem is not simply ownership of data. The important questions are:

• Is the dataset difficult to reproduce?
• Is the data continuously refreshed?
• Does the incumbent obtain data from a dominant platform?
• Can rivals obtain equivalent data?
• Does exclusive access prevent meaningful entry?
• Can data from one market be leveraged into another?

Competition concern

A vertically integrated platform may use data from its dominant service to create a superior generative AI system and then use that AI system to reinforce the original monopoly.

This produces a possible data → AI → distribution → data feedback loop.

5. Compute Concentration

Machine creativity is computationally intensive.

Training frontier models may require:

• advanced GPUs;
• AI accelerators;
• high-bandwidth memory;
• specialized networking;
• large-scale cloud infrastructure;
• energy;
• data-centre capacity.

Therefore, competition may be restricted if a small number of firms control essential computational inputs.

The FTC has expressly identified concentrated chip and cloud markets as potential competition concerns for generative AI.

Possible anticompetitive conduct

A dominant cloud provider might theoretically:

• reserve scarce compute for its affiliated AI model;
• impose discriminatory pricing;
• restrict interoperability;
• impose switching costs;
• require exclusivity;
• obtain commercially sensitive information from AI customers;
• bundle cloud infrastructure with its own AI services.

6. Foundation-Model Concentration

Foundation models exhibit potentially strong economies of scale.

A larger model can benefit from:

• more training data;
• more compute;
• greater user feedback;
• larger developer ecosystems;
• greater brand recognition;
• more applications;
• more revenue for further investment.

This can generate a self-reinforcing competitive advantage.

The CMA's foundation-model review specifically examined how the emerging foundation-model ecosystem could evolve and what competition risks could arise.

Competition-law issue

The crucial question is whether scale produces legitimate efficiencies or becomes a mechanism for excluding equally or potentially efficient competitors.

7. AI Distribution as a Bottleneck

Even if a rival develops a technically competitive creative model, it may be unable to reach users.

Distribution bottlenecks can include:

• mobile operating systems;
• app stores;
• search engines;
• browsers;
• social networks;
• cloud platforms;
• enterprise software;
• advertising networks.

For example, if a dominant mobile operating system gives its own AI assistant privileged access to device functions while restricting rival AI assistants, the operating-system position can be leveraged into the AI market.

This issue has become particularly concrete in the EU: in July 2026, the European Commission adopted measures requiring Google to provide competing AI services effective interoperability with relevant Android features.

8. Self-Preferencing of Machine Creativity

A vertically integrated company may operate:

platform + AI model + creative application + distribution channel.

It may then have incentives to favour its own AI-generated products.

Examples include:

• ranking its AI-generated content above competitors;
• giving its AI assistant preferred access to device functionality;
• displaying its AI-generated advertisements more prominently;
• prioritising its own AI coding tool;
• favouring its own AI image/video generator;
• using proprietary search data to improve its own AI system.

The EU's current DMA measures concerning Google's search data and Android AI interoperability demonstrate how regulators are addressing these kinds of ecosystem advantages.

9. Tying and Bundling

A dominant firm may potentially tie:

AI creative service + dominant platform service.

For example:

• cloud + AI model;
• smartphone OS + AI assistant;
• search + AI answer engine;
• office software + AI writing assistant;
• image-editing software + proprietary AI model.

The legal question is whether the bundle produces legitimate technical efficiencies or instead forecloses competitors.

10. Exclusive AI Partnerships

AI development increasingly involves partnerships between:

• cloud providers;
• foundation-model developers;
• chip suppliers;
• application developers;
• data providers.

The FTC's investigation of major AI/cloud partnerships examined issues including access to computing resources, engineering talent, switching costs, exclusivity and access to sensitive information.

Such arrangements may generate substantial efficiencies. However, competition analysis must examine whether they also:

• prevent rivals from obtaining compute;
• lock developers into one cloud;
• prevent competing models from accessing distribution;
• transfer sensitive competitive information;
• create artificial switching costs.

11. Algorithmic and AI-Assisted Collusion

Machine creativity platforms may compete through algorithms rather than human decision-makers.

Potential concerns include:

• common pricing algorithms;
• common recommendation systems;
• AI-generated advertising prices;
• automated bidding;
• coordinated licensing terms;
• algorithmic restrictions on creators;
• automated exclusion of rival platforms.

The legal difficulty is determining whether parallel algorithmic conduct results from independent optimisation or from an agreement, communication or coordinated strategy.

12. Network Effects

Creative AI platforms can develop strong network effects.

More users produce:

More prompts → more feedback → better model → more users → more developers → more applications → more data → better model.

This may make markets contestable initially but increasingly difficult to enter later.

Competition authorities therefore need to consider dynamic competition rather than only present market share.

13. Switching Costs and Lock-In

Creative professionals may invest heavily in:

• prompts;
• workflows;
• APIs;
• model fine-tuning;
• proprietary embeddings;
• stored projects;
• plugins;
• automation systems;
• team training.

Consequently, switching from one AI platform to another can become costly.

A dominant firm might increase switching costs through:

• proprietary file formats;
• non-portable models;
• API incompatibility;
• restrictions on exporting data;
• contractual exclusivity;
• technical barriers.

14. Data Portability and Interoperability

Interoperability is particularly important in machine creativity because creative workflows often involve several systems.

For example:

AI writing model → design software → cloud storage → publishing platform

If one company controls multiple layers, it may prevent competitors from interoperating effectively.

The EU's 2026 Android AI interoperability measures illustrate the growing importance of this issue. The Commission stated that competing AI services should receive effective access to relevant Android features comparable to Google's own services.

15. AI-Generated Content and Market Foreclosure

Machine creativity can compete directly with human-created products.

Suppose an AI platform controls:

1. a dominant search engine;
2. a dominant advertising platform;
3. a generative AI system.

It may begin replacing links to independent publishers with AI-generated answers.

The resulting concern is not necessarily that AI-generated content is inherently anticompetitive. Rather, the question is whether the dominant intermediary uses control over distribution to divert demand from competing content suppliers.

16. Six Major Case Laws and Their Relevance

Because dedicated reported judicial decisions specifically titled “machine creativity markets” remain limited, the most useful authorities are cases involving the underlying competition principles: platform leverage, data, tying, interoperability, exclusivity, innovation and self-preferencing.

Case 1 — Google Search (Shopping) Case

Google Search (Shopping), Case AT.39740, European Commission

Facts

The European Commission examined Google's practice of positioning and displaying its own comparison-shopping service more prominently than competing comparison-shopping services.

Competition principle

The case is important for self-preferencing and leveraging dominance.

Relevance to machine creativity

An analogous problem could arise where a dominant search or platform operator gives its own generative-AI creative service preferential:

• ranking;
• visibility;
• access to users;
• data;
• advertising opportunities.

The central question is whether control over the dominant platform is being used to disadvantage competing AI creative services.

Case 2 — Google Android

Google Android, Case AT.40099, European Commission

The Android proceedings concerned contractual arrangements involving Google Search, Chrome, Play Store, Android licensing and competing Android forks.

The case illustrates how control over an operating-system ecosystem can be used to influence adjacent markets.

The 2026 CJEU judgment concerning Google's Android appeal continued to address contractual restrictions, tying, exclusive pre-installation and exclusionary effects.

Relevance to machine creativity

The same logic can become relevant where:

AI assistant + operating system + app distribution

are controlled by the same company.

If the platform gives its own creative AI privileged access while limiting competitors, competition authorities can examine:

• foreclosure;
• tying;
• exclusive pre-installation;
• interoperability;
• leveraging.

Case 3 — CCI v Google: Android Mobile Devices

Competition Commission of India, Case No. 39 of 2018

The CCI found Google dominant in relevant Android-related markets and examined agreements concerning pre-installation, search distribution, Android forks and related ecosystem restrictions.

The CCI imposed a ₹1,337.76 crore penalty in 2022.

Key principle

The case demonstrates the importance of:

• ecosystem dominance;
• pre-installation;
• default status;
• network effects;
• restrictions on alternative operating systems;
• leveraging dominance between connected markets.

Application to machine creativity

An AI company controlling a major device or software ecosystem could potentially use:

default AI assistant + pre-installation + exclusive access to system functionality

to disadvantage competing creative AI services.

Case 4 — CCI: Google Play Store

CCI, Google Play Store proceedings, Case Nos. 07/2020, 14/2021 and 35/2021

The CCI examined Google's position in Android app-store distribution and its policies affecting app developers. It imposed a ₹936.44 crore penalty.

Competition principle

The case illustrates the importance of:

• platform dependency;
• indirect network effects;
• app distribution;
• access conditions;
• leveraging platform power.

Relevance to machine creativity

AI creative applications may increasingly depend upon:

• app stores;
• cloud marketplaces;
• AI model marketplaces;
• plugin stores.

If a dominant platform simultaneously operates its own AI application, discriminatory distribution conditions could raise competition concerns.

Case 5 — Matrimony.com v Google / CUTS v Google

CCI, Case Nos. 07/2012 and 30/2012

The CCI examined allegations concerning Google's search practices, including preferential treatment of its own vertical services.

The matters are particularly relevant to search bias and preferential placement.

Relevance to machine creativity

Generative AI changes the form of search from:

10 blue links

to:

one AI-generated answer.

This creates an even more important competition question:

Who controls which information or creative service is presented in the AI-generated answer?

A dominant AI intermediary could potentially determine:

• which creators are cited;
• which publishers are surfaced;
• which creative platforms receive traffic;
• which commercial services are recommended.

Case 6 — CCI v Meta / WhatsApp Privacy Policy

In re: WhatsApp LLC / Meta Platforms Inc., CCI Case No. 30 of 2021

In November 2024, the CCI found that Meta had abused its dominant position in the market for OTT messaging apps through the implementation of WhatsApp's 2021 privacy policy and its data-sharing implications, imposing a ₹213.14 crore penalty and behavioural directions.

Competition principle

The case is significant because it demonstrates that data-related conduct can be analysed through competition law, particularly where dominance in one market can facilitate expansion into another.

Relevance to machine creativity

For AI, data may be used for:

• model training;
• personalisation;
• recommendation;
• advertising;
• fine-tuning;
• reinforcement learning.

Thus:

dominant user platform → compulsory/expanded data collection → AI improvement → stronger AI service → stronger platform

may create a competition concern.

Case 7 — FTC v Facebook

FTC v Facebook, Inc.

The US proceedings concerning Facebook/Meta examined the use of platform power and strategies relating to potential competitive threats.

Competition principle

The broader significance lies in examining whether an incumbent can protect its market position by acquiring or restricting emerging competitive threats.

Machine-creativity relevance

Generative AI markets develop rapidly. An established platform may seek to acquire:

• promising AI laboratories;
• model developers;
• AI creative applications;
• specialist data companies;
• AI infrastructure firms.

Competition authorities therefore need to consider whether an acquisition removes a potential competitor rather than merely an existing competitor.

Case 8 — Qualcomm

FTC v Qualcomm Inc., 969 F.3d 974 (9th Cir. 2020)

The case concerned Qualcomm's licensing practices and its position in cellular modem technology.

Competition principle

The case is useful for analysing:

• vertical relationships;
• licensing;
• technology inputs;
• bargaining power;
• exclusionary conduct.

Relevance to machine creativity

AI markets may similarly contain upstream bottlenecks involving:

• GPUs;
• AI accelerators;
• model APIs;
• proprietary datasets;
• cloud computing.

A dominant upstream input supplier could potentially affect downstream AI creativity markets through discriminatory access or contractual restrictions.

17. Competition Concerns Specific to AI-Creative Outputs

A. Creative-Input Foreclosure

A dominant AI company may obtain exclusive access to:

• publishers' archives;
• image libraries;
• music catalogues;
• video repositories;
• professional datasets.

This could make entry more difficult.

B. Creator Lock-In

Creators may become dependent on one AI platform because their:

• prompts;
• style settings;
• model customisations;
• project histories;
• workflows

are not portable.

C. AI Marketplace Self-Preferencing

A marketplace could favour its own:

• AI-generated designs;
• AI-written advertisements;
• AI music;
• AI video;
• AI coding tools.

This resembles the broader self-preferencing concerns already examined in digital-platform competition law.

D. Exclusive Model Access

A cloud provider could potentially make certain advanced models available exclusively through its infrastructure.

Competition analysis would need to distinguish:

legitimate investment incentives

from

arrangements that substantially foreclose competing AI developers.

E. Predatory or Subsidised AI Pricing

Generative AI services may initially be supplied:

• free;
• below cost;
• with substantial credits;
• bundled into another service.

Low prices can benefit consumers and accelerate adoption. However, competition law may examine whether below-cost strategies are deliberately used to eliminate rivals and whether recoupment or other exclusionary effects are plausible.

18. Innovation Competition

Innovation is particularly important in machine creativity.

Traditional competition analysis may ask:

What is the price?

AI markets require additional questions:

What model architecture is available?

How rapidly are models improving?

Can rival developers experiment?

Can creators switch models?

Can new AI laboratories obtain compute and data?

Does the incumbent have incentives to suppress disruptive technologies?

A practice can therefore be problematic even where consumers currently receive AI services at a very low monetary price.

19. Dynamic Competition

Machine creativity markets are unusually dynamic.

Today's dominant model may be displaced by:

• a smaller model;
• an open-weight model;
• a specialised model;
• a multimodal model;
• an agentic system;
• a new architecture.

Therefore, regulators should consider:

Current competition

Who competes today?

Potential competition

Who could realistically enter?

Innovation competition

Who could develop the next generation of technology?

Ecosystem competition

Can developers move between platforms?

Input competition

Can rivals obtain comparable data and compute?

20. Essential-Facility-Type Concerns

Some machine-creativity inputs could potentially become economically indispensable.

Examples might include:

• uniquely valuable datasets;
• particular AI infrastructure;
• specialised model APIs;
• dominant creative marketplaces;
• indispensable distribution systems.

However, not every important input constitutes an essential facility.

The legal analysis normally requires careful examination of:

1. indispensability;
2. lack of realistic alternatives;
3. feasibility of duplication;
4. competitive foreclosure;
5. justification for refusal;
6. impact on downstream competition.

21. Competition and Open-Source AI

Open-source and open-weight AI can lower entry barriers.

They can permit:

• smaller developers to build applications;
• academic experimentation;
• independent model development;
• greater interoperability;
• reduced dependence on dominant providers.

However, open models can also be strategically controlled through:

• restrictive licences;
• proprietary hosting;
• preferential cloud arrangements;
• limited access to training infrastructure.

Thus, “open” does not automatically mean competitive, and proprietary does not automatically mean anticompetitive.

22. Remedies

Competition authorities could potentially consider several remedies depending on the infringement.

Structural remedies

• divestiture;
• separation of business units;
• restrictions on acquisitions.

Behavioural remedies

• non-discriminatory access;
• interoperability;
• data portability;
• prohibition of exclusivity;
• transparent ranking;
• restrictions on self-preferencing;
• fair API access.

Data remedies

• data access;
• data portability;
• interoperability;
• restrictions on cross-use of data.

Contractual remedies

• limits on exclusive arrangements;
• reduced switching costs;
• termination rights;
• multi-cloud arrangements.

Monitoring

AI markets may require continuing monitoring because competitive conditions can change extremely rapidly.

23. Indian Competition-Law Perspective

The Competition Act, 2002 provides particularly relevant tools through:

• Section 3 — anti-competitive agreements;
• Section 4 — abuse of dominant position;
• Section 5 — combinations;
• Section 19 — inquiry into agreements and dominant position;
• Section 26 — investigation;
• Section 27 — orders upon finding contravention.

The CCI's Google Android decision is particularly useful for machine-creativity analysis because it examined ecosystem leverage, pre-installation, network effects, default positions and exclusion of competing services.

The Meta/WhatsApp decision is equally important for the data–dominance relationship.

24. European Union Perspective

The EU approach combines traditional competition law with the Digital Markets Act.

The DMA is especially relevant because some machine-creativity problems may arise before conventional Article 102 litigation becomes practical.

In July 2026, the European Commission adopted binding measures relating to:

• interoperability for competing AI services on Android;
• access to Google Search data for third-party search services.

The stated objective was to allow competing AI services and search services to compete more effectively.

This demonstrates a shift toward ex ante regulation of ecosystem bottlenecks, rather than relying exclusively on lengthy abuse-of-dominance proceedings.

25. United States Perspective

US competition law can address machine-creativity markets through:

• Sherman Act §1;
• Sherman Act §2;
• Clayton Act;
• FTC Act §5;
• merger-control mechanisms.

The FTC's AI work has specifically examined the relationship between cloud providers and generative-AI developers, including investments and partnerships involving major companies.

This reflects an important US concern: whether financial and contractual relationships between major cloud providers and AI developers could influence access to essential AI inputs.

26. Key Competition-Law Tests for Machine Creativity

27. Important Doctrinal Distinction

It is important not to treat technological superiority as an antitrust violation.

A company may legitimately become successful because it has:

• a better model;
• better engineering;
• better training;
• more efficient compute;
• superior products;
• better creative outputs.

Competition law intervenes when market power is maintained or extended through anticompetitive conduct, rather than merely because an AI system is technologically superior.

This distinction is especially important in rapidly evolving machine-creativity markets.

28. Overall Legal Analysis

The central competition problem can be represented as:

CONTROL OF DATA
↓
CONTROL OF COMPUTE
↓
SUPERIOR FOUNDATION MODEL
↓
SUPERIOR CREATIVE APPLICATION
↓
CONTROL OF DISTRIBUTION
↓
MORE USERS
↓
MORE FEEDBACK AND DATA
↓
STRONGER AI MODEL

This feedback loop can create substantial competitive advantages.

The competition-law question is therefore not simply whether a firm possesses a large AI model. It is whether the firm can use control over one layer of the machine-creativity ecosystem to foreclose competition at another layer.

29. Conclusion

Machine creativity markets present a new form of multi-layered competition. The relevant competitive asset may no longer be merely intellectual property or physical infrastructure. It may be the combination of data, compute, models, distribution, users, feedback and creative ecosystems.

The principal competition concerns include:

1. training-data concentration;
2. compute and cloud concentration;
3. foundation-model dominance;
4. AI distribution bottlenecks;
5. self-preferencing;
6. tying and bundling;
7. exclusive AI partnerships;
8. creator and developer lock-in;
9. data leveraging;
10. algorithmic coordination;
11. killer acquisitions and potential-competition loss;
12. reduction of innovation competition.

The most relevant existing precedents—Google Shopping, Google Android, CCI's Google Android decision, CCI's Google Play Store proceedings, Matrimony.com/CUTS v Google, CCI v Meta/WhatsApp, FTC v Facebook, and FTC v Qualcomm—do not all concern generative AI itself. Their value lies in establishing competition-law principles that can be applied to the emerging structure of machine-creativity markets.

The emerging regulatory direction is increasingly toward preserving contestability, interoperability, access to critical inputs and innovation competition, while still allowing firms to obtain legitimate returns from investment in AI technology. The FTC has expressly warned that control over essential AI inputs can give firms outsized influence across downstream economic activity.