Competition Law in the Age of AI: Legal Frameworks and Enforcement Challenges
- UCL Law for All Society
- Jan 23
- 4 min read
By Andrea Berkovic
Artificial intelligence has emerged as a critical intersection point for competition law enforcement, raising novel legal questions that challenge traditional antitrust frameworks. From algorithmic pricing collusion to merger control of strategic AI partnerships, regulators worldwide are grappling with how to apply century-old statutes to cutting-edge technologies.
A fundamental legal question concerns who bears liability for anticompetitive conduct effectuated by algorithms. The prevailing view among enforcement agencies is that firms remain fully responsible for algorithmic decisions. This principle extends to situations where firms do not fully understand their algorithms’ decision-making processes, creating significant compliance challenges for complex machine learning systems whose behaviour may be opaque even to their developers.
Merger Control and Novel Partnership Structures
Many AI deals fall below traditional notification thresholds or involve unconventional structures that escape automatic review. In the UK, the Competition and Markets Authority’s investigation into the Microsoft-OpenAI relationship highlights regulators’ concerns that partnerships or acquisitions involving major technology firms could entrench market power and stifle innovation by smaller competitors. Microsoft invested nearly $14 billion in OpenAI beginning in 2019, securing exclusive cloud provider status and approximately 49 percent profit-sharing rights. The UK Competition and Markets Authority investigated whether this partnership constituted a relevant merger situation under the Enterprise Act 2002, examining whether Microsoft had acquired material influence over OpenAI's commercial policy. After a 15-month investigation, the CMA concluded in March 2025 that the partnership did not give rise to a relevant merger situation, though acknowledged the 2019 arrangements may have conferred material influence.
The concept of ‘acqui-hire’ arrangements has further tested merger control boundaries. In the Microsoft/Inflection case, the CMA asserted jurisdiction after Microsoft hired nearly all of Inflection's personnel and secured IP licensing rights, determining that Microsoft acquired the strategic capacity of Inflection's business, constituting acquisition of an enterprise under UK merger law, and thus a potential lessening of competition within a market or markets in the UK.
Algorithmic Pricing Under Section 1 of the Sherman Act
The application of Section 1 of the U.S Sherman Antitrust Act to algorithmic pricing has generated divergent judicial interpretations. The central legal question is whether competitors' use of common pricing algorithms, particularly those incorporating nonpublic competitive data, satisfies the agreement requirement.
In Duffy v. Yardi Systems, the US District Court for the Western District of Washington found that sharing sensitive, nonpublic commercial data with Yardi and subsequent use of its pricing recommendations could constitute a per se violation based on alleged horizontal price-fixing. By contrast, the Ninth Circuit in Gibson v. Cendyn Group rejected this theory, holding that independent agreements between multiple competitors and the same software vendor cannot constitute a Section 1 violation without proof of horizontal coordination among competitors themselves. These cases highlight the difficulty of applying traditional competition law concepts to technologically mediated pricing conduct.
The Department of Justice has taken an aggressive stance, arguing in March 2025 that algorithmic price fixing constitutes a per se antitrust violation. However, the 2025 RealPage settlement—which produced no judicial finding that algorithmic rent-setting violated the Sherman Act and left much of the company's operations intact—reflects practical enforcement difficulties.
Recognizing federal antitrust law limitations in addressing algorithmic collusion, several states have enacted targeted legislation. California, New York, and Connecticut have each enacted or amended state antitrust statutes to prohibit certain uses of algorithmic or ‘revenue management’ pricing tools. These state-level initiatives reflect frustration with the requirement under Section 1 of the Sherman Act that plaintiffs prove a horizontal agreement, and a legislative effort to regulate algorithmic collusion more broadly than federal law has yet to do.
The EU AI Act and Competition Law Interaction
The European Union's AI Act, which entered into force in August 2024, represents the first comprehensive regulatory framework for artificial intelligence. While drafted to apply without prejudice to competition law, the AI Act nonetheless affects antitrust analysis in several ways.
The AI Act extends investigative powers of national competition authorities by requiring each Member State to establish or designate market surveillance authorities with broad oversight functions. These authorities can access company data to conduct compliance checks, potentially uncovering information relevant to competition investigations. By expanding regulators’ investigatory powers, the Act also mandates transparency requirements and information-sharing obligations. This could inadvertently facilitate information exchange among competitors, particularly regarding training data, price monitoring or matching, algorithmic decision-making processes, and technical specifications.
The European Commission has raised concerns that limited access to critical AI inputs—such as large, high-quality datasets and specialized chips for neural network training—may be constrained by costly licensing agreements and supply bottlenecks. These infrastructure challenges, including restricted access to cloud computing resources, have become key considerations in the Commission’s merger control assessments, particularly for transactions involving GPU manufacturers or major cloud service providers. Such bottlenecks can entrench market power, hinder new entrants, and raise antitrust concerns by potentially foreclosing competition in emerging AI markets.
Thus, the application of competition law to artificial intelligence presents fundamental challenges for legal doctrine developed in contexts beyond the sphere of technology. Courts and agencies face difficult questions about agreement formation in the absence of direct communication, control attribution for autonomous systems, and jurisdictional scope for novel transaction structures. The legal landscape remains in flux, with divergent judicial interpretations, aggressive agency enforcement theories, and state-level legislative experimentation all contributing to considerable uncertainty.
Sources:
Edited by Artyom Timofeev
Comments