Competition Law and Competition Governance of Autonomous Knowledge Systems

1. Introduction

Autonomous Knowledge Systems (AKSs) may be understood as advanced AI systems capable of independently collecting, retrieving, ranking, synthesising, generating, updating, and sometimes acting upon knowledge on behalf of users. They include AI research agents, autonomous information assistants, AI search engines, enterprise knowledge agents, retrieval-augmented generation systems, autonomous decision-support systems, and agentic platforms that can interact with external databases and services.

Competition law becomes particularly important because an AKS can simultaneously operate as:

• a knowledge-access gateway;
• a search and recommendation intermediary;
• a data aggregator;
• an AI model or inference provider;
• a platform connecting users and information suppliers;
• an advertising or commercial referral intermediary; and
• an autonomous agent capable of making choices on behalf of users.

The Competition Commission of India (CCI) has specifically recognised AI as an emerging competition-policy area. Its 2025 Market Study on Artificial Intelligence and Competition examined AI ecosystems, emerging competition concerns, market structures and regulatory responses.

There is not yet a mature body of reported judgments specifically dealing with “autonomous knowledge systems” as a distinct antitrust market. Consequently, the existing jurisprudence on search engines, digital platforms, data, interoperability, self-preferencing, tying, exclusionary conduct and essential facilities provides the principal legal framework.

2. Meaning and Economic Structure of Autonomous Knowledge Systems

An AKS can be represented as:

Data → Foundation Model → Retrieval/Knowledge Layer → Ranking/Reasoning → Recommendation → Autonomous Action → Feedback/Data → Improved System

This creates several potential competitive bottlenecks.

A. Data bottleneck

A dominant AKS may possess:

• proprietary search indexes;
• user-query histories;
• proprietary datasets;
• behavioural data;
• enterprise knowledge repositories;
• copyrighted or licensed databases;
• interaction data;
• feedback data.

The larger the user base, the greater the quantity of interaction data available to improve the system.

This can create a data-network effect.

B. Compute bottleneck

Advanced autonomous systems may depend upon:

• specialised GPUs;
• cloud infrastructure;
• inference infrastructure;
• model-training capacity;
• energy-intensive data centres.

Control over these inputs may affect entry by smaller AI developers.

C. Knowledge-index bottleneck

An AKS that controls a large-scale knowledge index can potentially obtain a significant advantage over competing systems.

This issue has become particularly important in the Google search litigation. The U.S. Department of Justice reported that the 2025 remedies in the Google search case included access for qualified competitors to certain search-index and user-interaction data, as well as search-result syndication. The court expressly recognised the relevance of these assets to emerging GenAI competition.

D. Interface bottleneck

If consumers increasingly access the internet through one autonomous knowledge agent rather than individual websites, the agent may become a gateway to markets.

That produces an important competition question:

Who controls the ranking and selection process through which the autonomous system decides what information, business, product or service the user sees?

3. Applicable Competition-Law Framework

A. Relevant-market definition

Traditional markets may not adequately capture AKSs.

Possible relevant markets include:

1. AI-powered knowledge services;
2. general search;
3. AI-assisted search;
4. enterprise knowledge-management services;
5. foundation-model services;
6. AI inference services;
7. AI-agent platforms;
8. specialised information retrieval;
9. AI advertising/intermediation services.

Market definition may need to consider:

• substitutability;
• quality;
• accuracy;
• latency;
• privacy;
• interoperability;
• switching costs;
• network effects;
• multi-homing;
• data advantages;
• computational scale.

The relevant market may therefore be multi-layered rather than a single conventional product market.

4. Major Competition Concerns

4.1 Self-Preferencing

A dominant AKS may give preferential treatment to:

• its own AI-generated answers;
• its own search products;
• its own databases;
• its own applications;
• affiliated commercial services;
• its own advertising inventory.

For example, if an autonomous knowledge system answers:

“Which software is best for this task?”

and systematically gives its affiliated product preferential placement irrespective of objective relevance, competition concerns can arise.

The Google Shopping judgment is particularly important here.

5. Case Law 1 — Google Shopping

Google LLC and Alphabet Inc. v European Commission, Case C-48/22 P (2024)

The Court of Justice upheld the €2.4 billion fine concerning Google's favouring of its own comparison-shopping service over competing services. The Court treated the conduct as an abuse of dominance under Article 102 TFEU.

Principle

Dominance combined with discriminatory preferential treatment in a platform's ranking mechanism can constitute abusive conduct where the practice has exclusionary effects.

Application to AKSs

An autonomous knowledge system may determine:

• which source is cited;
• which answer appears first;
• which supplier is recommended;
• which database is searched;
• which commercial option is presented.

If the operator manipulates that autonomous ranking to favour affiliated businesses, Google Shopping provides an important analytical precedent.

The important distinction is that an AKS's ranking may be considerably more complex than a traditional search-results page because it may generate a synthesized answer rather than merely display links.

6. Case Law 2 — Google Search / United States v Google

United States and Plaintiff States v Google LLC

The U.S. search-monopolisation litigation is highly significant for autonomous knowledge systems.

The 2024 liability judgment found Google liable under Section 2 of the Sherman Act for maintaining monopolies in general search services and general search text advertising through exclusionary distribution agreements. In 2025, the court imposed remedies addressing distribution arrangements and requiring certain access to search data and syndication services.

The DOJ subsequently emphasised that the remedies were intended to prevent Google from using its existing search dominance to extend market power into GenAI.

Application to AKSs

This case demonstrates the danger of leveraging an established information gateway into a new AI market.

An incumbent controlling:

Search → Users → Data → AI knowledge system

could potentially use advantages obtained in the old market to protect or extend its position in the new one.

Relevant concerns include:

• exclusive defaults;
• preinstallation;
• preferential distribution;
• withholding important data;
• restrictive agreements with device manufacturers;
• tying AI products to dominant search products.

7. Case Law 3 — Microsoft v Commission

Microsoft Corp. v Commission, Case T-201/04 (2007)

The EU General Court essentially upheld findings that Microsoft had abused its dominant position by refusing to provide interoperability information to competitors and by tying Windows with Windows Media Player.

Principle

Interoperability can become a competition issue where a dominant undertaking controls an important technological interface and uses that control to restrict competitors.

Application to AKSs

Autonomous knowledge systems may need interoperability with:

• search engines;
• cloud platforms;
• enterprise databases;
• document repositories;
• payment systems;
• APIs;
• identity systems;
• competing AI models;
• external agents.

A dominant AKS provider could theoretically restrict API access or make interoperability technically inferior.

This could produce digital foreclosure.

8. Case Law 4 — Bronner

Oscar Bronner GmbH & Co. KG v Mediaprint, Case C-7/97

The Court of Justice established the restrictive conditions under which refusal by a dominant undertaking to provide access to an infrastructure can constitute abuse.

The case concerned access to a newspaper home-delivery network.

Importance for AKSs

The case provides the classic framework for essential-facility-type claims.

Suppose a dominant knowledge platform controls an infrastructure that is indispensable for effective competition, such as:

• a critical knowledge index;
• an essential API;
• a unique data repository;
• an indispensable technical interface.

A refusal to provide access would not automatically violate competition law. The stringent legal conditions associated with essential-facility doctrine must be examined.

Thus:

Large data ownership ≠ automatic essential facility.

Indispensability, elimination of effective competition and the other applicable legal requirements remain important.

9. Case Law 5 — Google Android

Google LLC and Alphabet Inc. v European Commission — Android

The EU Android litigation concerned Google's contractual arrangements concerning Android devices, including restrictions connected with application distribution and search.

The case illustrates the competition risks arising where a dominant ecosystem controls several interconnected layers.

Application to AKSs

An autonomous knowledge system may exist inside an ecosystem containing:

Operating System + Browser + Search + AI Assistant + App Store + Cloud + Advertising

The operator could potentially condition access to one layer upon acceptance of another.

Examples include:

• AI assistant + browser;
• AI assistant + search;
• AI model + cloud;
• AI agent + app store;
• enterprise AI + proprietary database.

The lesson from the Android jurisprudence is that ecosystem power can produce exclusionary effects even where the individual products are technically distinct.

In India, CCI likewise found Google dominant in important Android-related markets and imposed a ₹1,337.76 crore penalty in its 2022 Android decision.

10. Case Law 6 — Matrimony.com v Google

Matrimony.com Ltd. & CUTS v Google LLC & Others, CCI, Case Nos. 07 & 30 of 2012

CCI found Google to have abused its dominant position in online general web search and web-search advertising services.

The CCI specifically recognised that the design of a search-results page can constitute an important dimension of competition and noted Google's position as an important gateway to the internet.

Application to AKSs

This is particularly relevant to autonomous knowledge systems because an AKS can become an even more powerful gateway than a conventional search engine.

Instead of displaying ten links, it may provide:

one generated answer.

Consequently, the system's choice of sources can determine which competitors receive:

• visibility;
• traffic;
• customers;
• reputation;
• commercial opportunities.

Competition governance must therefore examine answer-generation architecture, not merely traditional search ranking.

11. Case Law 7 — Umar Javeed v Google

Umar Javeed & Others v Google LLC & Another, CCI Case No. 39/2018

CCI's Android decision in this matter concerned Google's conduct in the Android mobile-device ecosystem. The order was issued on 20 October 2022.

Significance

The case illustrates how competition concerns can arise from the combination of:

• operating-system dominance;
• app distribution;
• search;
• device manufacturers;
• default arrangements;
• ecosystem restrictions.

AKS relevance

An autonomous knowledge system could similarly become embedded across multiple layers:

Device → Operating System → Browser → Assistant → Search → Knowledge Agent → Commercial Transaction

The more layers controlled by one undertaking, the greater the potential for leveraging and foreclosure concerns.

12. Case Law 8 — Meta / WhatsApp Privacy Case

In Re: Updated Terms of Service and Privacy Policy for WhatsApp Users, CCI, 2024

CCI's 2024 decision concerning WhatsApp's 2021 privacy-policy update is particularly relevant to AI because data itself may constitute a competitive input.

CCI concluded that the policy involved an unfair condition and found concerns regarding sharing of WhatsApp data within the Meta group, including implications for competition in online advertising. It imposed a monetary penalty of ₹213.14 crore.

Application to AKSs

Autonomous knowledge systems require enormous quantities of data.

Data-related competition issues may therefore include:

• compulsory data collection;
• excessive data accumulation;
• combining datasets across services;
• restricting portability;
• denying rivals access to commercially significant data;
• using data from one market to strengthen another;
• feedback-loop advantages.

For an AKS:

More users → more queries → more interaction data → better system → more users

can create a powerful feedback loop.

13. Case Law 9 — Intel v Commission

Intel Corp. v Commission, Case C-413/14 P

The Court of Justice's judgment concerned loyalty rebates and exclusionary conduct under Article 102 TFEU. The Court required the effects of the conduct to be appropriately assessed where the undertaking argued that the conduct was incapable of restricting competition.

Application to AKSs

An AKS provider might offer:

• discounted API access;
• free AI inference;
• bundled cloud credits;
• preferential model pricing;
• rebates for exclusive deployment.

Such practices should not be analysed solely from their formal contractual structure.

The actual questions include:

• Can rivals realistically compete?
• Is the practice capable of foreclosing competitors?
• What proportion of demand is affected?
• Are competitors equally efficient?
• Are there objective efficiencies?

14. Case Law 10 — Servizio Elettrico Nazionale

Servizio Elettrico Nazionale and Others, Case C-377/20 (2022)

The Court of Justice examined exclusionary conduct by an incumbent undertaking in the liberalisation of the Italian electricity market. The Court addressed the importance of assessing whether conduct is capable of producing exclusionary effects and whether the undertaking used methods other than competition on the merits.

AKS relevance

The case is useful for analysing legacy advantages.

A company may possess advantages because it historically controlled:

• search;
• social networks;
• cloud infrastructure;
• operating systems;
• enterprise software;
• data repositories.

Competition law should distinguish legitimate innovation from using inherited market power to exclude emerging AI competitors.

15. Competition Concerns Specific to Autonomous Knowledge Systems

A. Knowledge-index foreclosure

A dominant firm may control the underlying index used by competing AI systems.

Potential conduct:

• refusing API access;
• imposing discriminatory access conditions;
• charging excessive access fees;
• rate-limiting competitors;
• providing inferior access to rivals.

B. Self-preferencing of generated answers

This is one of the most significant concerns.

Suppose an AI platform owns:

• the model;
• the search engine;
• the knowledge database;
• the advertising platform;
• a shopping service.

Its autonomous answer could systematically direct users toward its own ecosystem.

The legal question becomes:

Is the answer-generation mechanism being used as an exclusionary instrument?

Google Shopping provides an important analogy.

C. Algorithmic discrimination

AKSs may independently determine:

• rankings;
• recommendations;
• citations;
• supplier visibility;
• prices;
• commercial referrals.

Even without an employee manually making the decision, the operator can potentially be responsible for designing the system, setting objectives and maintaining the relevant commercial incentives.

16. Algorithmic Collusion

Autonomous knowledge systems may communicate with other automated systems.

For example:

AI Agent A ↔ AI Agent B ↔ Pricing System ↔ Marketplace

could create the possibility of coordinated pricing or market allocation.

Competition law must distinguish:

Legitimate parallel behaviour

from

Concerted conduct

and

Algorithmic implementation of an agreement.

The fundamental issue remains whether there is an agreement, concerted practice, or unilateral exclusionary conduct satisfying the relevant competition-law test.

17. Data Advantages and Feedback Loops

An AKS may generate an unusual competitive cycle:

Users

↓

Queries

↓

Training/feedback data

↓

Improved model

↓

Better answers

↓

More users

↓

More data

This creates a data-network effect.

A dominant undertaking might therefore acquire a competitive advantage that is difficult for new entrants to reproduce.

The Google search litigation is important because the U.S. court recognised user data and scale as important competitive inputs and the later remedies included certain data-access provisions for qualified rivals.

18. Tying and Bundling

An AKS provider could potentially bundle:

• AI assistant + operating system;
• AI assistant + browser;
• AI model + cloud;
• AI agent + search;
• AI knowledge service + enterprise software;
• AI assistant + advertising.

Competition authorities would need to determine:

1. whether the undertaking is dominant;
2. whether there are separate products;
3. whether customers are forced or strongly induced to take both;
4. whether competitors are foreclosed;
5. whether efficiencies justify the conduct.

The Microsoft and Android jurisprudence provides important foundations for this analysis.

19. Interoperability and Portability

Competition governance should encourage:

Data portability

Users should, where legally and technically appropriate, be able to move:

• prompts;
• personal knowledge bases;
• conversation history;
• agent configurations;
• preferences;
• enterprise information.

Model interoperability

Where commercially and technically feasible:

• APIs;
• model-switching;
• agent protocols;
• retrieval interfaces

can reduce switching costs.

Microsoft demonstrates why interoperability can become an antitrust remedy where a dominant technological ecosystem uses interoperability restrictions to disadvantage competitors.

20. Autonomous Agents as Gatekeepers

The traditional digital economy had:

Search engine → user → website

The emerging autonomous economy may become:

User → AI agent → market

The AI agent can potentially:

• search;
• compare;
• negotiate;
• purchase;
• book;
• communicate;
• select suppliers;
• recommend financial or professional services.

This transforms the agent into an economic intermediary.

Its ranking decisions can therefore influence entire downstream markets.

21. Competition Governance Model

A suitable governance framework can be divided into seven layers.

Layer 1 — Market-structure monitoring

Competition authorities should monitor:

• concentration;
• entry barriers;
• compute access;
• data access;
• model concentration;
• cloud dependence.

Layer 2 — Conduct regulation

Examine:

• self-preferencing;
• tying;
• exclusivity;
• discriminatory API access;
• predatory pricing;
• loyalty incentives;
• refusal to interoperate.

Layer 3 — Data governance

Monitor:

• data combination;
• data portability;
• data exclusivity;
• data-sharing arrangements;
• privacy as a competitive parameter.

Layer 4 — Algorithmic accountability

Authorities should be able to investigate:

• ranking objectives;
• training incentives;
• recommendation architecture;
• discriminatory outputs;
• model changes;
• automated commercial decisions.

Layer 5 — Merger control

AI acquisitions should be assessed for:

• nascent competition;
• data accumulation;
• access to compute;
• vertical foreclosure;
• interoperability;
• ecosystem expansion.

This is especially important where conventional turnover thresholds fail to capture the value of AI transactions.

22. Indian Competition-Law Position

The principal statutory provisions are found in the Competition Act, 2002, particularly:

• Section 3 — anti-competitive agreements;
• Section 4 — abuse of dominant position;
• Sections 5–6 — combinations;
• Section 19 — inquiry;
• Section 26 — investigation;
• Section 27 — orders after inquiry;
• Section 32 — conduct occurring outside India but having effects in India.

For AKSs, Section 4 may be particularly important because dominance can potentially arise from control over:

• AI platforms;
• digital ecosystems;
• data;
• search;
• cloud infrastructure;
• app distribution;
• enterprise knowledge systems.

CCI's 2025 AI market study expressly examined emerging AI competition concerns and possible regulatory responses, indicating that AI competition has become an established competition-policy subject in India.

23. Remedies for Anti-Competitive Autonomous Knowledge Systems

Possible remedies include:

Structural remedies

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

Behavioural remedies

• non-discrimination;
• prohibition of self-preferencing;
• access obligations;
• interoperability requirements;
• data portability.

Technical remedies

• API access;
• transparent ranking parameters;
• audit logs;
• independent testing;
• interoperability protocols.

Data remedies

• data-sharing under controlled conditions;
• portability;
• restrictions on cross-use;
• separation of datasets.

Procedural remedies

Competition authorities may require:

• algorithmic records;
• model-change documentation;
• preservation of decision logs;
• independent technical audits.

24. Special Problem: Explainability

Competition authorities ordinarily need to establish:

conduct → market power → foreclosure → competitive harm

But an autonomous AI system may produce outcomes through:

• billions of model parameters;
• machine-learning optimisation;
• reinforcement learning;
• continuously changing retrieval systems.

Consequently, traditional evidence may be inadequate.

Competition investigations may require:

• model documentation;
• training-data information;
• ranking logs;
• A/B testing records;
• API-access records;
• computational-resource information;
• internal strategy documents.

The Meta litigation concerning the Commission's information requests illustrates the importance of regulators being able to obtain relevant internal information when investigating complex digital-market conduct.

25. Autonomous Knowledge Systems and Consumer Welfare

Consumer welfare in AKS markets cannot be measured exclusively by monetary price because many systems are offered at zero monetary price.

Relevant competitive parameters include:

• accuracy;
• reliability;
• speed;
• privacy;
• transparency;
• choice;
• interoperability;
• innovation;
• hallucination/error rates;
• source diversity.

Thus:

Competition may occur through quality, privacy, accuracy and innovation even where the monetary price is zero.

26. Six Core Legal Principles Emerging from the Case Law

27. Key Case-Law List — At Least 6

1. Google LLC & Alphabet Inc. v European Commission (Google Shopping), C-48/22 P (CJEU, 2024) — self-preferencing and leveraging.
2. United States & Plaintiff States v Google LLC (U.S. District Court, D.D.C.) — search monopolisation, distribution agreements and AI-related remedies.
3. Microsoft Corp. v Commission, T-201/04 (General Court, 2007) — interoperability and tying.
4. Oscar Bronner GmbH & Co. KG v Mediaprint, C-7/97 (CJEU, 1998) — refusal of access and essential facilities.
5. Google Android litigation / Google and Alphabet v Commission — ecosystem restrictions, tying and leveraging.
6. Matrimony.com Ltd. & CUTS v Google LLC, CCI, Case Nos. 07 & 30 of 2012 — search bias and gateway power.
7. Umar Javeed & Others v Google LLC, CCI Case No. 39/2018 — Android ecosystem dominance and exclusionary practices.
8. Meta/WhatsApp Privacy Policy matter, CCI, 2024 — data, unfair conditions and leveraging.
9. Intel Corp. v Commission, C-413/14 P — exclusionary rebates and effects analysis.
10. Servizio Elettrico Nazionale and Others, C-377/20 — exclusionary effects and leveraging of inherited market advantages.

28. Conclusion

Competition governance of autonomous knowledge systems requires a shift from analysing individual software products to analysing AI ecosystems and control points.

The principal competition risks are:

Data concentration + compute concentration + knowledge-index control + autonomous ranking + ecosystem integration + interoperability restrictions + self-preferencing = potential competitive bottlenecks.

The existing jurisprudence does not yet create a separate legal doctrine called the “Autonomous Knowledge System Doctrine.” Instead, established principles concerning dominance, exclusionary conduct, self-preferencing, tying, interoperability, essential facilities, data advantages and leveraging can be applied to these systems.

The most important future competition question will therefore be:

Who controls the autonomous intermediary through which users obtain, evaluate and act upon knowledge?

If a small number of firms control that intermediary layer, competition law may need to ensure that their technological advantages do not become mechanisms for permanently excluding rival models, information providers, applications and downstream businesses.

CCI's dedicated 2025 AI competition study confirms the increasing importance of these issues in India, while the Google search remedies demonstrate that competition authorities are already considering how traditional digital-market power can affect the development of generative-AI markets.