However, in an increasingly digital economy, many firms are turning to algorithms and machine learning to optimise pricing, improve efficiency and respond faster to ever-changing market dynamics. Yet, what is often designed as an efficiency tool can, under certain conditions and circumstances, drift into the territory of inadvertent collusion.
Whilst this digital evolution challenges long-standing assumptions about intent, agreement and proof in competition enforcement and the law in general – across the globe, competition regulators are beginning to scrutinise algorithmic pricing tools and price signalling mechanisms that align competitors’ prices in ways that harm consumers.
The risk for businesses is clear – you do not need to meet your competitors in a smoke-filled room to collude. Increasingly, your software may do it for you.
Price signalling is not a new concept in competition law. It occurs when firms indirectly communicate their pricing intentions – through public statements, announcements, or even conduct that reduces strategic uncertainty and independent action.
Historically, regulators have treated such conduct as problematic where it acts as a substitute for explicit coordination.
From a South African perspective, specifically in accordance with section 4(1)(b)(i) of the Competition Act 89 of 1998 (as amended) (Competition Act), price fixing between competitors is per se unlawful, even in the absence of an explicit agreement. Section 4(1)(a) of the Competition Act also captures these kinds of “concerted practices” to the extent that they substantially lessen or prevent competition.
Thus, the challenge arises when pricing algorithms – now common in sectors like retail, aviation and digital intermediary platforms – start making those “signals” themselves.
Modern algorithms, particularly with the advent of artificial intelligence and machine learning, can:
Further, when multiple firms deploy similar algorithms, or use the same third-party software provider, the result can be coordinated outcomes without any conscious communication.
From a legal standpoint, this raises a crucial question – can firms be held liable if their algorithms ‘learn’ to collude?
Whilst most competition law frameworks – including South Africa’s – predate the conception of mediums such as algorithmic pricing and machine learning, regulators are rapidly adapting oversight mechanisms to meet this new reality.
Around the world, enforcement agencies are sending an increasingly consistent message: the use of automated systems does not absolve firms of responsibility for anticompetitive outcomes.
Since as early as 2015, the DOJ has made it clear that when competitors use algorithms to fix prices – even indirectly – this constitutes a per se antitrust violation (United States v. Topkins). The DOJ’s stance is that algorithms are merely extensions of human decision-making.
Similarly, the Federal Trade Commission has flagged the use of pricing bots and online data-scraping tools as potential vehicles for unlawful coordination.
Further, several national competition authorities, including those in Germany, France and the Netherlands, have initiated sectoral inquiries into algorithmic pricing in the retail, travel and consumer goods markets.
The CMA has urged companies to integrate algorithmic compliance systems, stressing that accountability rests squarely with the deploying firm. The CMA has identified risks across three categories, namely – collusion, exclusionary conduct and exploitative pricing – and emphasised that regulators are building the technical capacity to audit algorithms directly.
Viewed together, these developments mark a shift from mere awareness to enforcement preparedness. Regulators are no longer speculating about algorithmic collusion – they are actively mapping where it may occur and updating investigative techniques to detect and prosecute it.
Thus, the boundary between competitive intelligence and coordinated conduct is becoming increasingly blurred in algorithm-driven markets, and firms must understand where innovation ends, and infringement begins.
Accordingly, companies should be aware of the following ways in which algorithmic utility could inadvertently become collusive conduct:
In each case, the absence of human communication does not eliminate the presence of collusive effects, and regulators are becoming increasingly comfortable inferring theories of harm and concomitant liability from outcomes rather than intent.
The rise of algorithmic pricing represents both a remarkable efficiency opportunity and a growing compliance frontier. Whilst competition law principles remain grounded in human conduct, their application must now extend to the digital proxies that influence market outcomes.
Firms cannot afford to treat algorithms as black boxes – as accountability attaches not only to intent, but to effect.
As regulators in South Africa and abroad continue to refine their enforcement frameworks, companies would be well-advised to bring competition compliance into the heart of their digital strategies – embedding legal oversight alongside technological innovation.
In the age of machine learning, vigilance is no longer optional – it is a legal necessity.
