Forward to the Past: A Critical Note on the European Parliament’s Approach to Artificial Intelligence in Private International Law


On 20 October 2020, the European Parliament adopted – with a large margin – a resolution with recommendations to the Commission on a civil liability regime for artificial intelligence (AI). The text of this resolution is available here; on other issues of AI that are part of a larger regulatory package, see the Parliament’s press release here. The draft regulation (DR) proposed in the resolution is noteworthy from a choice-of-law perspective because it introduces new, specific conflicts rules for artificial intelligence (AI) (on the general issues of AI and PIL, see the conference report by Stefan Arnold here). With regard to substantive law, the draft regulation distinguishes between legally defined high-risk AI systems (Art. 4 DR) and other AI systems involving a lower risk (Art. 8 DR). For high-risk AI systems, the draft regulation would introduce an independent set of substantive rules providing for strict liability of the system’s operator (Art. 4 DR). Further provisions deal with the amount of compensation (Art. 5 DR), the extent of compensation (Art. 6 DR) and the limitation period (Art. 7 DR). The spatial scope of those autonomous rules on strict liability for high-risk AI systems is determined by Article 2 DR, which reads as follows:

“1.        This Regulation applies on the territory of the Union where a physical or virtual activity, device or process driven by an AI-system has caused harm or damage to the life, health, physical integrity of a natural person, to the property of a natural or legal person or has caused significant immaterial harm resulting in a verifiable economic loss.

  1. Any agreement between an operator of an AI-system and a natural or legal person who suffers harm or damage because of the AI-system, which circumvents or limits the rights and obligations set out in this Regulation, concluded before or after the harm or damage occurred, shall be deemed null and void as regards the rights and obligations laid down in this Regulation.
  2. This Regulation is without prejudice to any additional liability claims resulting from contractual relationships, as well as from regulations on product liability, consumer protection, anti-discrimination, labour and environmental protection between the operator and the natural or legal person who suffered harm or damage because of the AI-system and that may be brought against the operator under Union or national law.”

The unilateral conflicts rule found in Art. 2(1) DR would prevail over the Rome II Regulation on the law applicable to non-contractual relations pursuant to Art. 27 Rome II, which states that the Rome II Regulation shall not prejudice the application of provisions of EU law which, in relation to particular matters, lay down conflict-of-law rules relating to non-contractual obligations. Insofar, it must be noted that Art. 2(1) DR deviates considerably from the choice-of-law framework of Rome II. While Art. 2(1) DR reflects the lex loci damni approach enshrined as the general conflicts rule in the Rome II Regulation (Art. 4 Rome II), one must not overlook the fact that product liability is subject to a special conflicts rule, i.e. Art. 5 Rome II, which is considerably friendlier to the victim of a tort than the general conflicts rule. Recital 20 Rome II states that “[t]he conflict-of-law rule in matters of product liability should meet the objectives of fairly spreading the risks inherent in a modern high-technology society, protecting consumers’ health, stimulating innovation, securing undistorted competition and facilitating trade”. In order to achieve these purposes, the Rome II Regulation opts for a cascade of connections, starting with the law of the country in which the person sustaining the damage has his or her habitual residence when the damage occurred, provided that the product was marketed in that country (Art. 5(1)(a) Rome II). If that connection fails because the product was not marketed there, the law of the country in which the product was acquired governs, again provided that the product was marketed in this state (Art. 5(1)(b) Rome II). Finally, if that fails as well, the Regulation returns to the lex loci damni under Art. 5(1)(c) Rome II, if the product was marketed there. This cascade of connections is evidently influenced by the desire to protect the mobile consumer from being confronted with a law that may be purely accidental from his point of view because it has neither a relationship with the legal environment that he is accustomed to (his habitual residence) nor to the place where he decided to expose himself to the danger possibly emanating from the product (place of acquisition). The rule reflects the presumption that most consumers will be affected by a defective product in the country where they are habitually resident. Insofar, Art. 2(1) DR is, in comparison with the Rome II Regulation, friendlier to the operator of a high-risk AI system than to the consumer.

Even if one limits the comparison between Art. 2(1) DR and the Rome II Regulation to the latter’s general rule (Art. 4 Rome II), it is striking that the DR does not adopt familiar approaches that allow for deviating from a strict adherence to lex loci damni. Contrary to Art. 4(2) Rome II, where the person claimed to be liable and the person sustaining damage both have their habitual residence in the same country at the time when the damage occurs, Art. 2 DR does not allow to apply the law of that country. Moreover, an escape clause such as Art. 4(3) or Art. 5(2) Rome II is missing in Art. 2 DR. Finally yet importantly, Art. 2(2) DR bars any party autonomy with regard to strict liability for a high-risk AI system, which deviates strongly from the liberal approach found in Art. 14 Rome II.

Apart from the operator’s strict liability for high-risk AI systems, the draft regulation would introduce a fault-based liability rule for other AI systems (Art. 8 DR). In principle, the spatial scope of the latter liability rule would also be determined by Art. 2 DR as already described. However, unlike the comprehensive set of rules on strict liability for high-risk systems, the draft regulation’s model of fault-based liability is not completely autonomous. Rather, the latter type of liability contains important carve-outs regarding the amounts and the extent of compensation as well as the statute of limitations. Pursuant to Art. 9 DR, those issues are left to the domestic laws of the Member States. More precisely, Art. 9 DR provides that

“Civil liability claims brought in accordance with Article 8(1) shall be subject, in relation to limitation periods as well as the amounts and the extent of compensation, to the laws of the Member State in which the harm or damage occurred.”

Thus, we find a lex loci damni approach with regard to fault-based liability as well. Again, all the modern approaches codified in the Rome II Regulation – the cascade of connecting factors for product liability claims, the common habitual residence rule, the escape clause, and party autonomy – are strikingly absent from the draft regulation.

Moreover, the draft regulation, in principle, limits its personal scope to the liability of the operator alone (as legally defined in Art. 3(d)–(f) DR). Recital 9 of the resolution explains that the European Parliament “[c]onsiders that the existing fault-based tort law of the Member States offers in most cases a sufficient level of protection for persons that suffer harm caused by an interfering third party like a hacker or for persons whose property is damaged by such a third party, as the interference regularly constitutes a fault-based action; notes that only for specific cases, including those where the third party is untraceable or impecunious, does the addition of liability rules to complement existing national tort law seem necessary”. Thus, for third parties, the conflicts rules of Rome II would continue to apply.

At first impression, it seems rather strange that a regulation on a very modern technology – artificial intelligence – should deploy a conflicts approach that seems to have more in common with Joseph Beale’s First Restatement of the 1930’s than with the modern and differentiated set of conflicts rules codified by the EU itself at the beginning of the 21st century, i.e. the Rome II Regulation. While the European Parliament’s resolution, in its usual introductory part, diligently enumerates all EU regulations and directives dealing with substantive issues of liability, the Rome II Regulation is not mentioned once in the Recitals. One wonders whether the members of Parliament were aware of the European Union’s acquis in the field of private international law all. In sum, compared with Rome II, the conflicts approach of the draft regulation would be a regrettable step backwards. It remains to be seen how the relationship between the draft regulation and Rome II will be designed and fine-tuned in the further course of legislation.