EU Regulators Target AI Pricing Algorithms, With Serbian Law Adding Criminal Penalties

The Investigation That Put Algorithmic Pricing on the Regulator's Radar

A growing number of companies are using artificial intelligence and pricing algorithms to set the optimal price for their products. The software tracks demand, competitor moves, cost changes and inventory levels, then recommends or implements a price. What was once a powerful management tool is now drawing the attention of competition authorities on both sides of the English Channel — and beyond.

In March 2026, the United Kingdom's Competition and Markets Authority (CMA) opened an investigation into three international hotel chains — Hilton, IHG Hotels & Resorts and Marriott — and the property data firm CoStar, which operates a shared platform used by those competitors. The CMA suspects that commercially sensitive data was exchanged through the platform, enabling the hotels to align pricing without a formal agreement. The case is the most visible example of a broader shift: regulators are actively testing whether pricing algorithms, especially those that pool data from multiple competitors, amount to illegal coordination.

The European Commission had already signalled its new posture. At a July 2025 conference, it confirmed it was examining several potential cases involving algorithmic pricing across the EU, and by early 2026 multiple investigations were active. The core worry is that algorithms can mimic — and even strengthen — old-fashioned price fixing, but without an email trail or a smoked‑filled room.

The legal ripples are spreading well outside Brussels and London. In Serbia, the Law on Protection of Competition mirrors EU competition rules and expressly prohibits any tacit agreement or concerted practice that fixes prices. What sets Serbia apart is the criminal dimension: Article 229 of the Criminal Code punishes individuals who conclude a prohibited restrictive agreement with six months to five years in prison, plus a fine. If a pricing algorithm is found to constitute a restrictive agreement, employees and managers who authorised its use could face jail time.

Why Pricing Algorithms Are Now Seen as Collusion Machines

The CMA Hotel Probe and the Hub‑and‑Spoke Model

The investigation into Hilton, IHG, Marriott and CoStar centres on a pattern that competition lawyers call hub‑and‑spoke coordination. A single platform — the “hub” — collects and processes data from multiple competitors. By feeding each competitor the same information or recommendations, the platform can align pricing behaviour without the rivals ever communicating directly. The CMA’s statement that the probe reflects its commitment to stop new technologies harming fair competition signals that it views the platform, not just the hotels, as a source of collusion. If the case concludes with penalties, the appetite to pursue other shared‑platform business models will almost certainly grow.

Why EU Regulators Are Treating Algorithms as Collusion Tools

Before pricing algorithms became widespread, a competitor had to guess a rival’s costs and strategies. Aligning prices reliably required direct contact. Modern algorithms change that equation. Even when two companies use different software, algorithms designed for the same industry often track the same parameters — demand data, competitor price feeds, public reviews — and follow similar optimisation rules. The result is a heightened ability to predict how competitors will price, which, in turn, weakens the incentive to undercut. The European Commission’s concern is that algorithm‑driven “parallel” behaviour can push prices closer to a monopoly level than would occur in genuinely competitive conditions. Regulators are no longer searching just for a paper agreement; they are scrutinising the design of the algorithm itself, the data streams it consumes, and the degree to which those data sources are shared across competitors.

Serbian Law's Additional Sting: Criminal Liability

Serbia’s Law on Protection of Competition prohibits any explicit or tacit agreement, concerted practice, or decision of an association that fixes purchase or sale prices. The law explicitly covers indirect price‑fixing, which means a pricing algorithm that absorbs competitor data and pushes the market towards a higher common price could fall under the ban. The criminal code adds a sharp personal edge: a natural person who concludes a restrictive agreement inside a business entity faces imprisonment and a fine. If a regulator determines that a pricing algorithm constitutes such an agreement, the door opens to criminal charges against the individuals who decided to deploy it. Serbian courts and the national competition authority, the Commission for Protection of Competition, are likely to follow the interpretive paths set by EU institutions, making every enforcement decision in Brussels and London directly relevant to the domestic market.

How Companies Can Reduce the Legal Risks of Pricing Software

• Audit the algorithm’s data diet. If the software collects transaction or pricing data from a platform used by several competitors, the regulatory risk is highest. Companies should verify whether the algorithm pools data from multiple rivals into a single optimisation model.
• Choose — or configure — software that avoids cross‑competitor coordination. Insist on solutions that do not rely on real‑time competitor‑specific datasets that are also available to those competitors through the same provider.
• Document how the algorithm reaches a price. Record the parameters the software uses (input costs, own historical sales, public demand indices, etc.). In a regulatory proceeding, a documented, auditable process can demonstrate that the company took measures to prevent anti‑competitive outcomes.
• Establish internal legal oversight of pricing algorithms. Regular reviews by legal or compliance teams, especially before rolling out new pricing logic, can help catch designs that replicate the hallmarks of hub‑and‑spoke coordination.
• In Serbia, factor in personal criminal risk. Executives and managers should understand that involvement in a pricing algorithm later judged to be a restrictive agreement could trigger individual criminal liability, including imprisonment.