Introduction

Competition law aims to promote healthy competition in a country. It operates on the premise that the market should be regulated and bans anti-competitive agreements between firms. The emergence of artificial intelligence (AI) means that traditional competition laws may need reforms to broaden the threshold for anti-competitive behaviours to ensure that the vision of competition law, which is to form a market with perfect competition, can be realised.

In this research paper, the authors focus on traditional aspects of competition law, regulated by the European Commission (EC), and discuss how AI can have positive and negative implications in the region of competition. Topics the authors focus on are: AI in anti-competitive agreements (horizontal and vertical); abuse of dominance; and combinations (merger control). 

Anti-Competitive Agreements

Horizontal price fixing constitutes a significant infringement under European Union (EU) competition law. It arises when competitors at the same market level coordinate their pricing rather than competing independently. Such conduct undermines the primary objective of EU competition policy, which is to ensure that markets remain competitive, efficient and advantageous for consumers. The typical objectives of price fixing include increasing profits, stabilising prices and reducing competitive uncertainty. Due to its gravity, EU law classifies horizontal price fixing as a restriction of competition by object and enforces strict legal measures to prevent it. 

A defining feature of EU competition law is the ‘object’ standard. Under this principle, certain types of agreements are presumed to be anti-competitive without taking into account whether or not this actually affects the market. Price fixing falls into that category. Agreements that directly set prices or establish minimum pricing levels are automatically unlawful under this standard. The European Commission and national competition authorities do not need to demonstrate actual effects on competition in detail; the very nature of the conduct is considered sufficiently harmful. 

The concept of collusion under EU law is intentionally broad. It does not mandate a formal written contract between competitors. Explicit agreements are not needed to prove collusion since EU law also recognises that coordination can take informal forms. For example, EU law also captures informal coordination, known as concerted practices. These occur when firms deliberately coordinate their behaviour instead of making independent decisions, even without directly communicating or forming a formal agreement. This can include informal gentleman’s agreements (unwritten understandings between firms) or repeated behaviour that, together with other evidence, suggests coordination instead of coincidence. Purely parallel conduct alone is not sufficient to prove collusion under EU law and needs corroborative evidence. 

Another important aspect is the treatment of information exchange. Sharing commercially sensitive information between competitors is particularly problematic under EU law. This includes future pricing intentions, cost structures, strategic plans or pricing models. Such exchanges reduce strategic uncertainty in the market, making it easier for firms to anticipate and align with each other’s behaviour. In some cases, depending on the nature and context of the exchange, information exchange can be treated as a restriction by object, but it can also be assessed as a restriction by effect. 

Traditionally, enforcement of these rules relied heavily on evidence of human communication. Cases are often built on emails, meetings, phone calls or documents showing explicit coordination. However, the emergence of artificial intelligence in pricing strategies has significantly complicated this framework. AI systems are increasingly used by firms to set prices dynamically based on large datasets, including competitor pricing, demand patterns and consumer behaviour. While these systems are designed to maximise efficiency and profit, they can also lead to unintended coordination across firms. 

An example of an instance in which automated systems led to horizontal price-fixing is the European Union antitrust case Eturas UAB and Others v. Lietuvos Respublikos konkurencijos taryba. In this 2016 case, more than 30 independent travel agencies used the same online booking website. The platform administrator changed the software code to cap all online consumer discounts at 3% automatically. Even though the agencies never met or talked directly to plan this change, their individual booking systems were subject to the platform’s restriction. This created an artificial limit on discounting across the market, reducing price competition. 

Following the Eturas judgment, it is important to note that the case did not result in any criminal convictions. EU competition law is primarily enforced through civil and administrative proceedings rather than criminal prosecution. Instead of prison sentences, the Lithuanian Competition Council imposed administrative fines on the platform operator and participating travel agencies. Following the guidance provided by the Court of Justice of the European Union (CJEU), the Supreme Administrative Court of Lithuania upheld most of the fines, while reducing some to reflect each firm’s level of involvement. Several agencies had the charges against them dismissed because there was insufficient evidence that they had actually seen the internal message announcing the discount restriction. One agency benefitted from leniency after cooperating with the competition authority and providing evidence that helped uncover the conduct. 

More importantly, the Eturas judgment established several legal principles that continue to shape the application of EU competition law to digital markets. First, the Court confirmed that merely receiving an electronic message through an online platform is not enough to prove participation in a cartel. Competition authorities must first establish that an undertaking was actually aware of the communication, reflecting the fundamental principle of the presumption of innocence. Second, once knowledge of the communication has been established, the Court held that continued use of the platform without objection can allow a presumption of participation in a concerted practice. This presumption can be rebutted if the company demonstrates that it publicly distanced itself from the conduct, objected to the platform administrator, reported the behaviour to the competition authorities or otherwise acted inconsistently with the anti-competitive restriction. Finally, the judgment made clear that businesses cannot avoid liability simply because the coordination was facilitated by a third-party digital platform rather than through direct communication with competitors. Companies remain responsible for ensuring that the technology they use complies with competition law. 

Although Eturas involved relatively basic software, its significance extends far beyond the travel industry because it illustrates how technology can facilitate coordination between competitors. Today’s artificial intelligence systems are considerably more advanced, with many businesses using machine learning algorithms to analyse competitors’ prices, predict consumer demand and automatically adjust prices in real time. These systems improve efficiency and allow firms to respond quickly to market changes, often without direct human involvement in individual pricing decisions. 

However, these same capabilities have raised concerns about algorithmic collusion. Unlike traditional price-fixing, algorithmic collusion may occur without an explicit agreement or any communication between competitors. If several firms use AI systems designed to maximise profits, the algorithms may independently learn that maintaining higher prices is more profitable than engaging in aggressive price competition. This creates the possibility that markets become less competitive even though there is no evidence of a traditional cartel. Such outcomes challenge the foundations of Article 101 of the Treaty of the Functioning of the European Union (TFEU), which has historically relied on identifying some form of agreement or concerted practice between undertakings. 

This has sparked considerable debate among competition law scholars and regulators over whether existing legal frameworks are sufficient to address AI-driven markets. The European Commission has generally maintained that Article 101 TFEU is broad enough to capture algorithmic collusion where there is evidence that businesses knowingly adopted or relied on systems that facilitated coordination. At the same time, proving liability has become significantly more complex. Traditional evidence such as emails, meetings or telephone calls may still be relevant, but authorities increasingly also analyse algorithms, software design and pricing data. Competition authorities must therefore develop greater technical expertise to distinguish between lawful algorithmic pricing, which reacts to market conditions, and unlawful coordination that restricts competition. 

Looking ahead, many experts argue that competition law will need to evolve alongside advances in artificial intelligence. Greater transparency requirements for pricing algorithms, increased oversight of automated decision-making systems and clearer guidance on corporate responsibility for AI-generated outcomes are frequently proposed as ways to strengthen enforcement. At the same time, regulators must avoid discouraging innovation, as AI offers substantial benefits through improved efficiency, lower operating costs and more responsive pricing. The challenge for EU competition law is therefore to maintain effective protection against horizontal price fixing while ensuring that businesses can continue to develop and use artificial intelligence in ways that promote, rather than undermine, competitive markets.

Vertical Anti-Competitive Agreements and AI

A vertical supply chain setup could, for example, include a manufacturing company and a distribution company. The part of the supply chain in which the manufacturing company occupies is the part where value is added to raw materials or components so as to transform them into sellable products that satisfy the market demand; whereas, the distribution company behaves as the vital link between manufacturing companies and retail facilities by purchasing the sellable products from manufacturing companies in bulk, storing them in large storage facilities and transporting them to the retail stores. 

Having understood a vertical supply chain setup, vertical collusion would mean collusion in a vertical supply chain. A vertical collusion is a secret agreement between businesses at different levels of the same supply chain to limit competition so as to reap mutual benefits. 

An example of a vertical collusion setup can be in a supply chain with a manufacturer and a distributor. If a manufacturing company is one of three manufacturing companies that operate within a sparsely populated pastoral region, and the distribution company manages to convince all three manufacturing companies into signing an exclusive dealing contract (a type of vertical collusion), this would mean that, for those three manufacturing companies, they can thereby only sell their goods to that one distribution company. For the distribution company, they have artificially attained regional market dominance (overshadowing any other distribution company), and they can abuse this dominant position and charge retailers large amounts of money that do not reflect market demand nor supply. The retailer may then be forced to raise the price of the purchased goods for consumers, which means that the vertical collusion between the manufacturers and the distribution company is ultimately to the consumers’ detriment. 

Another example of vertical collusion is vertical price fixing. This is when a manufacturer forces retailers to sell a product at a minimum price. This reduces retailer competition as retailers can no longer freely compete for discounts. This means that products remain at their very high prices at all times, and price sensitive customers will be left with few options to choose from. This is because they cannot shop around to find the best value for money if all retailers sell a specific product at the same price. In some cases, vertical price fixing can be positive as it can protect brand reputation, and for products that use precious, finite resources, it can aid in reducing over production and product waste. However, vertical price fixing becomes collusive and illegal when it is used to facilitate or enforce horizontal collusion among competitors rather than simply managing a brand; this is also known as a hub and spoke setup. 

While vertical collusion is harder to spot than horizontal collusion because supply chain partners have legitimate reasons to communicate and share data unlike competitors on the same level of the supply chain, discovering businesses participating in vertical collusion can be facilitated through the use of machine learning. Machine learning is a subset of AI that enables systems to learn from data and improve their performance over time. These AI models can be trained to analyse pricing patterns and market demand of goods and spot any mismatch between the two that could hint at vertical collusion which can then be investigated. Moreover, the Competition and Markets Authority (CMA) operates through a leniency policy under which businesses and individuals that provide evidence of vertical collusive activity can benefit from a reduction in or, in some circumstances, complete immunity from penalties. Thus, AI could potentially be used in enforcement activities and help in identifying vertical collusion and ultimately tackling it.

Abuse of Dominance: How dominant firms use AI to maintain market control

While self-preferencing is one of the principal ways in which dominant firms may abuse their market power, it represents only one stage of a broader competition law framework. Competition law first considers whether a firm has acquired a dominant market position before examining whether that dominance has been abused through exclusionary or anti-competitive conduct. Given the context of the rapidly evolving AI industry, artificial intelligence plays a role at both stages; it can contribute to the creation and maintenance of dominant market positions, while also introducing new avenues through which dominant firms may reinforce or abuse that market power. Consequently, understanding AI’s impact requires examining not only how it shapes market structure, but also how it influences the conduct of firms once dominance has been established.

DOMINANCE

Market dominance is achieved once a single firm holds a substantial share of the market, giving it the power to strongly sway prices, product standards and supply, and set other market conditions without pushback. By itself, market dominance is not illegal under global competition laws; rather, the company has the responsibility of ensuring it does not abuse its power to stifle competition. As a result, competition law seeks to prevent dominant firms from abusing market power in order to protect consumers while encouraging innovation and motivating competitors. However, the emergence of artificial intelligence has transformed the ways in which firms can acquire and reinforce market power. Markets become naturally concentrated through the increasing reliance on control of data, computing infrastructure and network effects, creating new challenges in preventing anti-competitive conduct. 

For instance, when large firms already collect substantial quantities of user data, AI models allow for the possession of extensive real-world user data that new entrants struggle to replicate and thus compete with. This is evident in industry leader Tesla Inc., which collects billions of miles of driving data from its vehicles, used to continuously improve autonomous driving technologies. As competitors are unable to employ comparable volumes of data, they face large obstacles in developing AI models of similar capabilities. As a result, data functions as the barrier that reinforces existing power within concentrated markets. 

Furthermore, the computing costs behind training advanced AI models easily lean towards tens to hundreds of millions of dollars while consuming staggering amounts of electricity. The significant amount of resources required, including the thousands of necessary specialised GPUs, creates rigid financial and operational barriers. As noted by the Federal Trade Commission in the FTC Staff Report on AI Partnerships & Investments 6(b) Study (2025), the scarce partnerships between large cloud service providers and artificial intelligence developer partners may shift control access to these resources towards market leaders and disadvantage competitors. Already established firms gain a clear structural advantage, further cementing their position as industry leaders. 

Additionally, as more users interact with the artificial intelligence of a firm, the AI systems learn specific user preferences and improve exponentially in quality as they obtain more data. Unfortunately, this creates a self-reinforcing loop where more users are attracted to the dominant firm as smaller competitors struggle to maintain a competitive alternative with their smaller amounts of data.

ABUSE

As firms attain dominant market positions, their artificial intelligence capabilities enable them to reinforce that dominance through exclusionary practices (a type of abuse). While it is true that some strategies may enhance innovation, others may trigger several anti-competitive risks. As previously discussed, self-preferencing represents a key mechanism through which firms can preserve control by giving less visibility to competitors. This occurs when dominant platforms can manipulate AI outputs or algorithms to artificially favour their own services. However, beyond self-preferencing lie several other abusive practices, such as bundling, tying and ecosystem lock-in. 

The common anti-competitive strategy of bundling and tying is where a firm combines multiple products or services. Although painted as a method to offer better value to consumers, companies are regulated under competition law to prevent firms with a near monopoly of a certain product from using their power to force consumers to purchase less desired products or shut out smaller competitors from the tied market. Recently, more cases have been rising where AI features are bundled into existing dominant products. This is exemplified in Microsoft’s aggressive integration of Copilot into Microsoft 365. The bundle has drawn mass scrutiny and has led to several global antitrust probes. Specifically, it led to investigations from regulators such as The Italy Competition Authority (AGCM) for raising Microsoft 365 prices after bundling AI tools without consumer consent and the U.S. Federal Trade Commission (FTC) for allegedly stifling competing AI models. As consumers already rely on the essential marketed product, the bundled AI services receive immediate exposure, consequently eliminating other competitors and reinforcing the market position of the dominant firm. 

Another related strategy includes the practice of ecosystem lock-in, where a company’s interconnected products embed users so profoundly in their ecosystem to the extent that switching to a competitor may be expensive or inconvenient, essentially “trapping” the consumer. Apple Inc. famously blends its products through both hardware and software, where several key features only work seamlessly within the Apple ecosystem. As artificial intelligence becomes increasingly integrated across systems, the cost of switching to rival providers will most likely discourage consumers from exploring other services, allowing dominant firms to strengthen their market control. 

While these anti-competitive strategies demonstrate how dominant firms can reinforce their market position, antitrust authorities thoroughly review these schemes to ensure they comply with competition law. Due to this, the European Union formally introduced the Digital Markets Act (DMA) in 2022, a landmark law that prevents dominant digital companies from enforcing an ecosystem on consumers.

Combinations: Traditional Combinations and Their Regulation Under EU Competition Law

Traditional combinations are generally understood as arrangements made between two or more entities involving the merging of resources, operations or in some cases, commercial activities. These combinations can occur between direct competitors/entities (horizontal combinations) or between businesses at different stages of the supply chain (vertical combinations). These standard agreements made between certain groups are often subjected to antitrust reviews when specific financial thresholds are crossed. 

Horizontal combinations happen when businesses operating at the same level of the market, such as two competing companies, work together or merge. These combinations can reduce competition because there are fewer rivals. In contrast, vertical combinations involve businesses at different stages of the supply chain, such as a manufacturer and a retailer. These agreements can improve efficiency and reduce costs, but they may also make it harder for other businesses to compete. Such combinations are often reviewed by the European Commission, the legal body responsible for presiding over general competition law, but specifically through its Director-General for Competition, which investigates antitrust violations and enforces merger controls between certain parties in order to ensure that the market is not negatively impacted by the joining of two companies. However, with the rise of artificial intelligence, it is important to note that traditional merger control thresholds may fail to measure a startup’s true market value. As AI firms possess non-tangible assets such as data and algorithms, regulators often underestimate their competitive significance and true potential. 

With the rapid development of the AI industry, queries are presented on whether or not merger-control rules ought to expand to encompass new technology. Numerous studies have established that to adequately address AI-driven markets, merger control must be expanded from its traditional form to accommodate even non-conventional or non-traditional metrics when analysing mergers and acquisitions. A contrasting view contends that an expanded merger control framework would stifle corporate growth and economic dynamism. Director of Competition Policy, Dirk Auer, argues that an expansion to encompass “non-controlling partnership” risks legal uncertainty and overreach. Proponents argue that predicting future competitors is speculative in the fast-changing AI markets. Moreover, researchers worry that over-enforcement may hinder innovation and investment. In this view, to encourage the development of the AI market, merger control must remain grounded in measurable criteria to preserve competition and avoid vague legal standards. 

While both perspectives raise valid concerns, at the current stage of the growing AI industry, it is premature to draw a definitive conclusion regarding the expansion of traditional merger controls. Nevertheless, a more balanced approach may be required where traditional merger control is retained, coupled with greater scrutiny of acquisitions surrounding AI inputs.

Conclusion

Overall, while artificial intelligence introduces several challenges to the enforcement of competition law, it is important to recognise that these implications are not inherently negative. Rather, AI also brings substantial benefits, particularly in relation to increased efficiency, improved pricing mechanisms and greater innovation across markets. Consequently, the future of competition law relies heavily on a balance of regulations behind artificial intelligence and flexible enforcement so as not to discourage the pro-competitive benefits that AI can offer.