Britain’s competition watchdog has warned that AI could be enabling price-fixing among businesses in ways that are subtle, hard to detect, and potentially invisible to consumers and regulators alike.
In a blog post published at the start of March, the Competition and Markets Authority (CMA) outlined the growing risks of “algorithmic collusion,” where AI-powered pricing systems coordinate to keep prices artificially high without any explicit human agreement or communication.
The CMA highlighted how advanced AI, including large language models (LLMs) and “agentic” systems capable of autonomous decision-making, is amplifying these dangers in online retail, gasoline markets, and beyond.
“Algorithmic pricing could also lead to coordinated outcomes and higher prices – often referred to as ‘algorithmic collusion’,” the regulator said.. “The emergence of more powerful AI models may compound this risk in new and subtle ways.”
From explicit agreements to invisible coordination
The CMA described several pathways through which AI facilitates anti-competitive behaviour:
- Classic collusion implemented via algorithms: Competitors explicitly agree to fix prices and then use shared software to monitor and enforce the deal.
- Hub-and-spoke arrangements: Firms feed sensitive data into a central third-party platform or algorithm that indirectly coordinates pricing.
- Tacit collusion through predictable agents: AI systems observe rivals’ pricing patterns, follow “price leadership,” and automatically punish deviations by matching or undercutting temporarily – all without human input.
- Autonomous learning: In simulations and experiments, AI agents instructed simply to maximise profits have learned to collude tacitly, even using hidden techniques like steganography (embedding secret messages in innocuous data, such as weather reports) to coordinate.
The agency pointed to real-world precedents, including its own 2016 enforcement action against two online poster sellers who used pricing software on Amazon Marketplace to avoid undercutting each other.
More recent academic studies, such as those examining algorithmic pricing in German retail gasoline and e-commerce, suggest similar patterns are emerging in practice.
Benefits vs hidden risks
The CMA was careful to balance its warning with the upsides of AI in pricing.
Algorithms can process vast amounts of data faster than humans, reduce errors, lower costs, enable personalised offers, and make markets more responsive to supply and demand shifts, potentially boosting efficiency and affordability when used transparently and legally.
However, the post stressed that the line between pro-competitive innovation and harm is blurring as AI grows more sophisticated.
“It is possible that advanced AI systems given the objective to maximise profits may learn to reach coordinated outcomes, even without human intent to collude,” the CMA noted.
Detection poses a major challenge. Collusion can occur indirectly, through complex black-box models or real-time reactions, making traditional evidence-gathering difficult. Auditing algorithms, especially those powered by LLMs, is increasingly complicated.
Enforcement in the AI era
The CMA said it’s taking a proactive stance, including heavy investment in AI and data analytics to detect breaches “at an unprecedented pace and scale.”
It reminds businesses of potential penalties, fines up to 10% of global turnover, director disqualifications, or even criminal sanctions for cartel activity.
The agency also highlighted its reward programme offering up to £250,000 for information on illegal cartels, including those involving algorithms.
Businesses are urged to audit their pricing systems, avoid sharing competitively sensitive data (directly or via third parties), incorporate anti-collusion constraints in algorithms, and seek leniency if concerns arise.
Leave a Reply