Pál Szilágyi: The human factor in EU competition law: The liability of non-human companies for antitrust infringements – a quick take. Nr. 2026/6.

In June 2026, the Argentine government proposed a new corporate category — the non-human corporation — operated by AI agents or robots, endowed with limited liability, with human shareholders optional. Conceived as a deregulatory device for walling risk inside an autonomous entity, it poses a question EU competition law has not yet confronted: can such an entity be held liable for an antitrust infringement, and if so, who pays? This paper argues that the answer turns on the functional concept of the undertaking rather than on corporate form. Because Article 101 TFEU looks past legal status and national registration, a non-human corporation offering goods or services on a market is an undertaking like any other, and its Argentine pedigree affords no immunity — least of all given the extraterritorial reach of EU competition law. The decisive questions are ones of attribution. Where a human or corporate parent controls the entity, the Akzo presumption of decisive influence pierces the limited-liability wall and exposes the controllers. Where the entity is genuinely ownerless and run by non-deterministic autonomous agents, the paper identifies a real but bounded enforcement gap: true algorithmic autonomy may negate the concurrence of wills that Article 101 requires (the conscious-parallelism limit of Wood Pulp), while non-determinism defeats the human-awareness pivot on which attribution doctrines such as Eturas rest. Liability, if it attaches at all, then falls on a thinly capitalised and likely judgment-proof entity. EU competition law thus remains applicable to non-human corporations in principle; what strains is its anthropocentric attribution architecture — designed for human cartelists — at precisely the point where autonomy and ownerlessness coincide.

The working paper is available for download here or at SSRN.