How can artificial intelligence do business in the European Union’s internal market? My recently published paper shows how artificial intelligence could be granted legal personality under the laws of Union member states. The paper then discusses the resulting effects in the internal market of the European Union. These effects, according to the paper, come down to ‘free movement of algorithms’.

Essentially, artificial intelligence that enjoys legal personality under a member state’s national law can invoke the freedom of establishment laid down in the European Union’s founding treaties to have its personality recognized in other member states. According to the case law of the Court of Justice of the European Union on freedom of establishment – Daily Mail, Centros, Ueberseering – the mutual recognition of personhood is mandatory with regard to artificially intelligent companies. These are companies ‘housing’ artificial intelligence, which have been established by means of a mechanism first proposed by U.S. scholar Shawn Bayern (for a simple explanation of how this mechanism works, see my previous blog post). By contrast, e-personhood, which the European Parliament recently put forward in a report dated 31 May 2016 on Civil Law Rules on Robotics, confronts serious difficulties. It will likely be unfeasible, since the European Union would not have the prerequisite legislative power to introduce it, and the consensus necessary to adopt the required legislative measure may prove elusive.

Under internal market law, the member states are largely precluded from adopting measures infringing on the free movement of artificially intelligent persons, save in narrowly circumscribed circumstances. Accordingly, measures against artificially intelligent persons may be adopted exclusively when such persons are not subject to any kind of human control. Indeed, in such a situation, a risk exists that the law – and in particular criminal law – would not be fully enforceable against them. Hence, targeted measures ensuring effective enforcement of the law are compliant with internal market requirements.

However, these measures must ensure that a natural person can be held criminally liable, when an uncontrolled artificially intelligent person commits a criminal offence. National law may thus require artificially intelligent persons to indicate a responsible natural person, e. g. on a letterhead or in a register.

By contrast, denying to recognize the legal personality of an artificially intelligent person would not be a lawful measure, because it would violate the freedom of establishment, would not be suitable to ensure effective enforcement of the law and would, in any case, go beyond what that objective requires.

Thomas Burri is an Assistant Professor of International Law and European Law at the University of St Gallen.