This is arguably the most difficult frontier for antitrust law. It challenges some of the core assumptions on which competition enforcement has been built for decades. Unlike traditional cartels, AI systems can produce collusive outcomes without any explicit agreement or communication between firms.
In classical antitrust controls, the enforcement authorities look for evidence of coordination: messages, meetings, or signaling. They can literally turn any office upside-down in search of physical evidence. However AI changes this rule. Pricing algorithms trained to maximize profit may independently arrive at the same strategy of raising prices and maintaining them because deviation triggers immediate “retaliation” from competitors’ systems. It’s like the systems “battling” themselves and the war is very tactical and fierce. Therefore, there are cases where no tangible inter-firm communication is needed[1].
This creates a form of “algorithmic cartel[2]”:
When one firm raises prices, others follow almost instantly, not through classic coordination, but through rapid response to market signals. The result mimics collusion, even if no human ever agreed to it. Algorithmic pricing can produce outcomes resembling collusion, even in the absence of an explicit agreement among human decision-makers. Emerging legal controversies, such as the RealPage litigation, and empirical evidence from retail gasoline markets illustrate the potential for pricing algorithms to weaken competition[3]. However, these cases differ in the extent to which they involve information-sharing arrangements, human coordination, or independently learning algorithms.
The situation becomes even more complex when advanced machine learning models are involved. Many such systems operate as “black boxes[4]”, meaning even their own creators cannot fully explain how specific decisions are made.
If two independent AI agents learn that maintaining higher prices leads to better long-term profits, they may converge on that strategy autonomously (and some times simultaneously), taking into account that algorithms approximate demand conditions using large-scale consumer data. This raises some critical legal dilemmas: Who is responsible for the cartel-like outcome? Can intent exist without human direction? How do we prove collusion when there is no communication?
In many jurisdictions, proving intent or agreement is central to antitrust enforcement. Let alone the hardship of the authorities to actually prove it. But with AI, harmful market outcomes can emerge without either and nobody can guarantee that businesses won’t actually exploit this as a “window” of opportunity, a grey zone, in order to speculate without being accountable. It’s a challenge and the institutions try to adapt to it, some, (especially EU) already use effects-based analysis and in any case, Courts can infer (albeit with difficulty) coordination from behavioral patterns and market structure.
This tension is forcing regulators to reconsider the foundations of competition law. Some experts argue enforcement may need to shift away from proving intent and toward evaluating outcomes, such as sustained supra-competitive pricing or reduced consumer welfare[5].
However, this approach carries risks. Penalizing outcomes alone could discourage legitimate competitive strategies or innovation in pricing technologies.
Strategic Implications for Businesses:
For companies deploying AI pricing systems, this is a real operational risk. Firms must now consider embedding specific guardrails to prevent anti-competitive behavior, monitoring algorithmic decisions for unintended convergence and, as with other AI-generated decisions, ensuring human oversight in high-impact pricing strategies.
It’s no longer a world where companies and humans compete each other in market terrms. We now have to deal with algorithms competing with algorithms making the line between optimization and collusion no longer clear. The issue is probabilistic, emergent, and increasingly difficult to regulate.
[1] Song Lixun, AI Algorithmic Pricing for Online Platforms: A Literature Review
Advances in Economics, Management and Political Sciences, 2025
[2] https://betterfinance.eu/article/algorithm-cartels/
[3] Sokler B. et al. (2025), https://www.mintz.com/insights-center/viewpoints/2191/2025-12-01-last-years-rent-realpage-reaches-settlement-agreement#:~:text=Following%20a%20year%2Dlong%20federal,commercial%20revenue%20management%20software%20and
[4] Hassija, V., Chamola, V., Mahapatra, A. et al. Interpreting Black-Box Models: A Review on Explainable Artificial Intelligence. Cogn Comput 16, 45–74 (2024). https://doi.org/10.1007/s12559-023-10179-8
[5] Abdullah S. (2025), https://sites.lsa.umich.edu/mje/2025/11/17/algorithmic-pricing-colluding-without-breaking-the-rules
