1. Introduction

Artificial Intelligence (AI) is increasingly becoming part of competitive markets. Firms use AI for pricing, advertising, product recommendations, hiring, logistics, demand forecasting, procurement, credit assessment, customer service, search, content distribution, and strategic decision-making.

AI can strengthen competition by reducing costs, improving innovation and enabling new firms to enter markets. At the same time, it can create significant competition-law risks where AI systems:

• facilitate price coordination;
• exchange competitively sensitive information;
• strengthen dominance;
• discriminate between trading partners;
• engage in self-preferencing;
• foreclose competitors;
• create switching costs;
• exploit data advantages;
• facilitate mergers or acquisitions that eliminate future competition; or
• reproduce anticompetitive strategies at unprecedented speed.

The central competition-law challenge is therefore to distinguish AI-enabled independent competition from AI-enabled coordination or exclusion.

2. AI as a Competition-Law Issue

AI affects competition through several interconnected mechanisms.

Major areas include:

1. Algorithmic pricing
2. Algorithmic collusion
3. AI-enabled information exchange
4. AI and abuse of dominance
5. AI-driven self-preferencing
6. AI and tying/bundling
7. Data advantages
8. AI mergers and acquisitions
9. AI infrastructure and compute access
10. Generative AI and platform competition
11. AI interoperability and access
12. AI-related vertical restraints
13. AI and innovation competition
14. AI procurement and public-sector markets

3. AI and Article 101 TFEU / Cartel Law

Article 101 TFEU prohibits agreements and concerted practices that restrict competition, subject to the applicable legal framework.

AI creates several potential forms of coordination.

Traditional cartel

Competitors communicate directly:

Firm A → Firm B → agreed price.

Algorithm-mediated cartel

Competitors communicate through a pricing system:

Firm A → algorithm ← Firm B.

Autonomous algorithmic coordination

The firms may not expressly communicate with one another, but algorithms continuously monitor market conditions and respond predictably.

This creates a difficult question:

When does algorithmic parallelism become legally relevant coordination rather than lawful independent conduct?

The answer depends upon evidence of the parties' conduct, knowledge, communications, design choices and the applicable legal test.

4. Algorithmic Pricing

AI can determine prices dynamically by analysing:

• competitor prices;
• demand;
• inventory;
• customer behaviour;
• historical sales;
• location;
• time;
• purchasing patterns.

Dynamic pricing itself is not unlawful.

Indeed, it can create substantial efficiencies.

The competition concern arises where algorithms are used to:

• implement an existing cartel;
• facilitate communication between competitors;
• monitor compliance with a cartel;
• exchange commercially sensitive information;
• deliberately coordinate prices; or
• make deviation from coordinated conduct easier to detect.

5. Algorithmic Collusion

Algorithmic collusion can be divided into four broad categories.

Type 1 — Human agreement + AI implementation

Competitors agree to fix prices and use software to implement the agreement.

This is the most straightforward situation.

Type 2 — Common algorithm

Competitors use the same third-party pricing system.

The competition analysis depends upon how the system operates and what information it uses.

Type 3 — Algorithm-mediated coordination

A platform or intermediary communicates competitively sensitive information between competitors.

Type 4 — Autonomous learning

Independent AI systems learn that particular pricing strategies produce higher profits and repeatedly respond to each other's behaviour.

The fourth category raises the most difficult unresolved questions.

6. AI and Information Exchange

Information is one of the most important competitive assets in AI markets.

AI systems can process huge quantities of information concerning:

• prices;
• costs;
• customers;
• capacity;
• inventory;
• demand;
• production;
• future commercial strategies.

An information exchange that would be difficult or expensive to conduct manually can become instantaneous through APIs and AI systems.

Consequently, competition authorities may need to investigate not merely what information was exchanged, but:

• how frequently;
• at what level of detail;
• between whom;
• through what system;
• whether it was historical or current;
• whether it concerned future conduct; and
• whether competitors could identify the source.

7. AI and Abuse of Dominance

AI can strengthen an already dominant firm's market position.

Potential exclusionary strategies include:

• discriminatory access to AI infrastructure;
• denial of interoperability;
• self-preferencing;
• tying AI services to another dominant product;
• exclusive access to important data;
• discriminatory API access;
• predatory or exclusionary pricing;
• leveraging dominance from one market into another.

The relevant legal provisions depend upon the jurisdiction.

In the European Union, Article 102 TFEU is particularly important.

In India, Section 4 of the Competition Act, 2002 addresses abuse of dominant position.

8. AI and Data Advantages

AI systems frequently depend on large datasets.

A firm with access to substantially more:

• consumer data;
• transaction data;
• behavioural information;
• training data;
• proprietary feedback;

may potentially obtain competitive advantages.

However, possession of extensive data is not automatically an antitrust violation.

The legal analysis generally requires consideration of:

• market definition;
• dominance;
• substitutability;
• barriers to entry;
• data uniqueness;
• replicability;
• network effects;
• access conditions; and
• actual or likely competitive effects.