Heightened antitrust scrutiny of algorithms and AI

Competition authorities are increasingly focused on the competitive risks posed by algorithms and artificial intelligence. In recent public remarks, the European Commission (Commission) has highlighted pricing algorithms as an area of concern, noting their potential to facilitate coordination between competitors. Similar themes have emerged in the UK, where the Competition and Markets Authority (CMA) has emphasised that while algorithms and AI can generate substantial efficiencies and consumer benefits, they may also enable novel forms of collusion. In a recent blog post, the CMA explained that it is actively screening for algorithm‑driven coordination and set out expectations for the practical steps companies should take to manage these risks.

For further background, please see our earlier publications on artificial intelligence and competition: AI: redefining the boundaries of competition (law), AI-enhanced competition enforcement in the EU and the second edition of the book Digital Competition Law in Europe.

CMA blog post

In its blog post, the CMA has been careful to stress that algorithmic pricing itself is not a recent innovation. Automated pricing tools have been used for many years across a variety of sectors, including aviation, hospitality and retail. This continuity is reflected in its earlier analytical work, most notably the CMA’s 2018 study into pricing algorithms and its 2021 market study examining the potential negative effects of algorithms on competition and consumers.

What has changed is the scale, sophistication and pervasiveness of these tools. Contemporary pricing systems can ingest highly granular data, operate continuously in real time, and increasingly rely on advanced machine‑learning techniques and large language models. At the same time, the cost and accessibility of predictive technologies have fallen sharply. As a result, businesses can now deploy powerful tools that shape (or fully automate) key commercial decisions, including pricing, with limited human intervention.

A separate CMA analysis has also examined so‑called “agentic AI”: autonomous software agents that optimise pricing or other strategic parameters on behalf of firms. The CMA’s research and analysis identifies a distinct competition risk where agents used by competing firms interact in ways that dampen competitive intensity, a phenomenon often described as “agentic collusion”.

Examples of enforcement in practice

Enforcement practice across jurisdictions demonstrates that the use of algorithms does not insulate firms from liability. On the contrary, authorities have repeatedly made clear that companies remain responsible for the conduct of their automated systems, even where decision‑making is partially delegated to software or where senior management claims limited insight into how an algorithm functions.

Algorithmic pricing and competition law: managing exposure

The introduction of pricing software that relies on algorithms or artificial intelligence does not reduce a company’s responsibility under competition law. Liability cannot be avoided by pointing to third‑party vendors, automated decision‑making, or a lack of insight into the technical workings of a system. Businesses are expected to take ownership of how pricing outcomes are generated, including understanding the role of data inputs, the logic that connects those inputs to outputs, and whether similar tools are used by competitors. These considerations are particularly important where existing pricing systems are gradually supplemented with AI‑based features, potentially increasing risk without a clear moment of reassessment.

Close attention should be paid to the information processed by pricing tools and to the conditions under which that information is shared or accessed. The use of non‑public, commercially sensitive data requires special care, especially where the same software provider services multiple market participants. In such cases, even indirect mechanisms of information signalling or alignment may raise competition concerns.

Independent decision‑making remains a core requirement. Pricing software must support, rather than replace, autonomous commercial judgment. To that end, companies should implement effective internal controls, backed by well‑defined policies, supervision of day‑to‑day use, and training programs focused on the specific risks posed by algorithmic and AI‑assisted pricing. Even when software appears technically compliant, employees must not coordinate with competitors on how tools are configured or deployed. This principle applies not only to automated price‑setting systems, but also to tools that track or benchmark market behaviour, which, despite being less sensitive, may still enable practices such as resale price maintenance.

Risk‑mitigation measures in practice

A structured compliance approach is recommended, which may include the following elements.

Tool selection and onboarding

Before deploying algorithmic pricing or AI‑enabled solutions, businesses should undertake thorough pre‑implementation assessments as part of their procurement processes. This includes examining how the model operates, where training and operational data originate, and whether outputs could reflect or reinforce competitors’ pricing strategies. Beyond the tool’s stated purpose, companies should consider how its capabilities could expand over time, including the potential for more autonomous decision‑making. Where tools are commonly used by actual or potential competitors, seeking legal guidance prior to adoption may be prudent.

Operational controls and governance

Clear rules should regulate how pricing tools are configured and used internally. Companies should retain documentation on system settings, data inputs, and any human intervention in pricing outcomes. Significant changes to algorithms or functionality should be subject to internal review. Particular restraint is warranted when considering whether to input non‑public or sensitive information, supported by internal processes to assess data sensitivity and, where necessary, apply safeguards such as aggregation or delayed use.

Review and testing mechanisms

Ongoing oversight is essential. Regular reviews should examine not only outcomes, but also the integrity of underlying data, access rights, override mechanisms, and employee interaction with the system. Where AI tools incorporate language‑based models, behavioural or linguistic stress‑testing may help identify unintended risks, and technical guardrails should be implemented where possible.

Awareness and training

Employees involved in pricing decisions should receive focused training on the competition law implications of algorithmic tools and on the risks of sharing or indirectly exchanging sensitive information, including through platforms, consultants, or software providers. Maintaining awareness is critical to ensuring that technological efficiency does not come at the expense of legal compliance.

Conclusion

Loyens & Loeff closely monitors the enforcement actions of competition rules in the digital sector. The actions of the competition authorities across jurisdictions reflect the increased focus on addressing competition concerns relating to the use of algorithms and AI, and ensuring fair competition in rapidly evolving technology markets.   

Contact

If you have any questions or would like to explore the implications of these developments for your business, please feel free to get in touch with one of the advisers mentioned below.