Försäkringsskyldighet i svensk rätt – En utredning mot bakgrund av Europaparlamentets rekommendation om obligatorisk ansvarsförsäkring för AI

Sammanfattning: The EU is currently drafting a legal framework regarding liability for AI with the purpose of promoting the rollout of trustworthy AI, while ensuring effective compensation for potential victims. As a part of this, the European Parliament on 20 October 2020 recommended compulsory liability insurance for high-risk AI. Despite this, the succeeding proposal for a directive on AI-liability lacks such provisions. Instead, it prescribes a review five years ensuing the end of the transposition period, anew assessing the matter of mandatory insurance. Based on the methods of legal-dogmatic research and legal analysis, the essay aims to analyse and evaluate the proposal of compulsory insurance through the perspective of Swedish insurance law. The material is predominately composed of law, legislative history and legal literature. Compulsory liability insurance has been considered justified within Swedish law when there is a particular need for protection. This because of for example the collective of potential victims, the characteristics of the damages alternatively the area of business and lack of other options for compensation. The proposal appears to align with this, especially considering the hazardous character of high-risk AI and the possibly large damages followed by a risk of insolvency. Compulsory insurance furthermore requires available, adequate insurance options, which the market currently appears to lack in respect of high-risk AI. A recurring question is whether, and if so how, the insurance industry will be willing to insure high-risk AI and thereby undertaking the responsibility of governing high-risk AI and its risks. Principles within Swedish insurance law state that insurers are free to decide if and how they want to insure a risk or client. Lack of available insurance options can hinder the effectiveness of the compulsory insurance in regard of victim compensation. Regarding efficient victim compensation the following aspects must also be taken into account: lack of duty for insurers to insure high-risk AI, insurance exclusions and limits, the insurers possibilities to object a direct action taken by the victim, and systems for supervision along with sanctions. Liability insurance acts as a security for potential victims and improves chances of receiving compensation. With compulsory liability insurance, several for the victim favorable provisions in the Insurance Contracts Act become applicable, regarding for example direct action against the insurer and statutory limitation. Liability insurance also provides a sense of security for the obliged AI-operator, which can foster the rollout of AI in Europe. However, making such insurance compulsory may negatively impede the rollout because of higher costs and the insurance industry becoming forced to act as a gatekeeper regarding who is given access to the market. On the other hand, this presents an opportunity to control and govern the market of high-risk AI to a certain extent, thus seeing to the trustworthiness of the high-risk AI. In summary, it is established that the adequacy and success of the proposed compulsory insurance in regard of Swedish law, is dependent on the insurance industry’s willingness to insure high-risk AI and how the law stating the compulsion is constructed. This regarding for example supervision and sanctions, whether insurers should be obliged to contract high-risk AI and limitations of the insurers right to object against a direct action taken by the victim.