This morning I have participated in the CIRSF parallel sessión regarding the topic “Digitalizing investment finance” directed by Prof.. Luis Morais, in CMVM Annual Conference 2021, Public conference in conjunction with the 46th IOSCO Annual Meeting presenting my paper “Civil liability derived from the use of artificial intelligence and its insurance”. A summary of my intervention is as follows.
1. The spiral dynamics of reciprocal increase civil liability derived from the use of artificial intelligence and its insurance.
1.1. Civil liability derived from the use of AI and its insurance move according to a spiral dynamic of reciprocal increase that leads to the fact that:
a) Tthe quantitative and qualitative increase of civil liability derived from the use of AI produces a parallel increase in its underwriting
b) And the latter incentivizes liability claims by moving a potentially infinite dynamic.
1.2. In particular, the civil liability of the operators of AI systems and their insurance show, in recent times and more intensely in the immediate future, an interaction in the form of reciprocal interference that is directly proportional since, the greater the number of claims, the greater the need of the civil liability insurance. This situation will favor claims as users or consumers become aware of the existence of the solvent assets of the insurer to cover damages suffered as a result of the use of AI systems.
1.3. Today, we can see the contrast observed between a certain underdevelopment of the «claim culture» in the world of AI and the notable development of the uses of said AI. This contrast allows us to venture an exponential growth of civil liability claims -in judicial and extrajudicial seat- of consumers versus operators of AI systems in the near future.
1.4. It is more than foreseeable that civil liability claims against artificial intelligence system operators will increase in number and amount both for two kinds of reasons:
a) General reasons, such as the so-called “claim culture” that is widespread in developed countries with respect to economic activities in general and the professions in particular.
b) Specific reasons, such as the growing digitization of the economy and, in particular, of fintech, with the massive use of algorithms. This growing digitization will generate a greater risk of loss according to the “function + obligation + responsibility equation”.
2. Civil liability derived from the use of artificial intelligence.
2.1. Civil liability derived from the use of AI is based on some assumptions of the specific regulation of civil liability in the field of artificial intelligence (AI) systems that are based on the principles of a responsible artificial intelligence (RAI) in the EU.
2.2. These regulatory frameworks are projected:
a) In general, within the EU, in the Proposal for a Regulation of the European Parliament and of the Council on civil liability for the operation of artificial intelligence systems.
b) In particular, in Spain, in the civil liability of the promoter of controlled test spaces (“regulatory sandboxes”) established, in current Spanish law, by Act 7/2020, for the digital transformation of the financial system.
2.3. Two basic types of civil liability for the use of AI can be distinguished according to whether it derives from civil or criminal offenses. That is:
a) The civil liability of the operators of AI systems may derive from civil offenses, in the form of contractual or extra-contractual liability, in accordance with arts. 1101 and 1902, respectively, of the Civil Code;
b) The civil liability of the operators of AI systems may derive from criminal offenses, in the form of civil liability derived from a crime, a crime in accordance with art. 116 of the Penal Code. Regarding the latter, the direct civil liability of the insurer is established in article 117 of the Penal Code stands out.
2.4. In both cases, the common classic requirements must be proven to claim civil liability derived from the use of AI: first, the damage action; second, the damage caused in the form of consequential damage or loss of profit; and, third, the effective causal relationship between both. That is, the requirement of civil liability to the operators of AI systems requires proof of the concurrence of three essential requirements:
a) First, regarding the first requirement of the negligent or malicious action of the operator of AI systems, failures can be of two types:
a.1) Intentional failures, legally qualifiable as malicious or seriously negligent.
a.2) Unintentional or accidental failures, which may come from slight negligence or force majeure or acts of God.
In turn, both failures can come from the insured AI operator, its employees or a third party that interferes with the operation of such systems (for example, after the marketing of a product that incorporates AI factors).
Each type of failure will produce -or not- the birth of the corresponding civil liability debt and will be attributable to the respective causative party. In case of doubt and in defense of the user, operates the principle of solidarity of the agents of the AI systems.
b) Second, regarding the second requirement of the accreditation of patrimonial damage, this damage must be characterized as real (that its existence can be proven), be economically assessable and constitute a physical or mental injury or material or non-pecuniary damage.
The examination of the jurisprudence in this type of professional or business liability insurance allows us to venture that a significant part of the claims will be dismissed because the actors will not prove the property damage suffered. Without this being an obstacle to recall that, on some occasions, the courts have admitted as compensable damage the claim of reparation in natura by equivalence and have also condemned to compensate for non-pecuniary damage.
c) The third requirement consist of the accreditation of an effective causal relationship between the negligent action of the operator of AI systems and the patrimonial damage suffered by the user who claims.
Regarding this third requirement, we can predict, on the basis of the examination of the jurisprudence issued on this type of professional or business liability insurance, that a significant part of the liability claims of the operators of AI systems will be dismissed for lack of accreditation of that relationship of causality by the claimants.
3. The insurance of the civil liability of the operators of artificial intelligence systems.
3.1. Insurance is shown as the ideal instrument for preventive coverage at partial cost of the civil liability of the operators of AI systems for the benefit of the consumer.
3.2. The civil liability insurance of the operators of AI systems will be governed in Spain by two kinds of regulations:
a) Specific regulations that are applicable in the case of mandatory insurance of high-risk artificial intelligence systems (in the future).
b) Law of Insurance Contract, which will be applied with imperative character as a general rule or device if it is an insurance for large risks that meets the requirements of art. 11 of the Act 20/2015, of July 14, on the management, supervision and solvency of insurance and reinsurance entities (Ley 20/2015, de 14 de julio, de ordenación, supervisión y solvencia de las entidades aseguradoras y reaseguradoras, LOSSEAR).
The insurer must comply with the rules of conduct –especially regarding documentation and information- imposed by the LOSSEAR and the Royal Decree 1060/2015, of November 20, on the management, supervision and solvency of insurance and reinsurance entities (Real Decreto 1060/2015, de 20 de noviembre, de ordenación, supervisión y solvencia de las entidades aseguradoras y reaseguradoras, ROSSEAR).
3.3. The civil liability insurance of AI system operators will be carried out by means of a civil liability insurance that will present three essential characteristics:
a) First, it will be a mandatory insurance for high-risk artificial intelligence systems and voluntary for the rest of the systems of artificial intelligence.
b) Second, it will be a business or professional liability insurance.
c) And, third, it may be structured as an individual or group insurance.
3.4. Characteristics of the civil liability insurance of the operators of AI systems affect the following aspects:
a) Parties to the insurance contract.
b) Insurance regime.
c) Substantial delimitation of the risk covered by reference to AI systems.
d) Temporal delimitation of the risk covered by reference to claims filed against the insured AI system operator (“claims made” clauses).
e) Legal defense of the insured AI system operators against the claim of the injured user
f) Direct action of the user of a damaged AI system or his heirs against the insurer of the operator.
3.5. The critical points of the civil liability insurance of the operators of AI systems will be:
a) Material delimitation of the coverage, with the inclusions and exclusions agreed in the policy.
b) Temporal delimitation of the coverage, with the precision of the covered claims.
c) Legal defense regime of the defendant operator.
d) Direct action of the injured user against the insurer.
3.6. The adequate design of the civil liability insurance for the operators of AI systems should produce a double and reciprocal benefit:
a) For the AI operators, whose assets will be kept unharmed from the claims of the injured users.
b) For users, who will have the preventive and additional guarantee provided by the insurer.
Additional note: This paper has been extracted from the following studies: Tapia Hermida, A.J., “La Responsabilidad Civil derivada del uso de la Inteligencia Artificial y su Aseguramiento”, Revista Ibero-Latinoamericana de Seguros, Vol. 30, Núm. 54, 2021, pp. 107-146; Idem, “La responsabilidad civil derivada del uso de la inteligencia artificial y su aseguramiento”, Revista de la Asociación Española de Abogados Especializados en Responsabilidad Civil y Seguro n.º 76 (2020), pp. 79 a 104, ISSN: 1887-7001; Idem, “Decálogo de la inteligencia artificial ética y responsable en la Unión Europea”, La Ley Unión Europea n.º 87, Sección Doctrina, 23 pp. ISSN: 2255-551X, Diario La Ley, 30 de diciembre de 2020. ISSN: 1989-6913; and Idem “Decálogo de la inteligencia artificial ética y responsable en la Unión Europea, Diario La Ley, n.º 9749, 4 de diciembre de 2020. ISSN: 1989-6913.