AI and Legal Liability: How to Sue Artificial Intelligence and Obtain Compensation

It is here and increasingly permeating both personal and business life, but the speed at which everything is happening has somewhat caught lawmakers off guard, who are only now becoming aware that the field of artificial intelligence needs legal regulation. There is no doubt that this will be a priority in the coming years, but the European Union has already begun with several directives, currently in the form of proposals. Thus, the system for claiming compensation for individuals who have suffered damage from the use of artificial intelligence systems (AI systems) could soon be significantly eased thanks to two additional directives proposed by the European Commission in September 2022.

Namely, in the three-part regulatory system for preventing damage and regulating artificial intelligence in the European Union, there should – along with the Regulation on Artificial Intelligence – be a Directive on the Adaptation of Rules on Non-Contractual Civil Liability with regard to Artificial Intelligence (AI Liability Directive) and an amended and expanded Directive on Liability for Defective Products (Updated Product Liability Directive). Both will impose certain obligations on business entities and many institutions, so although these are currently only proposals, companies should already take steps that could prepare them for this regulation. Of course, the legal system will also need to adapt before these directives are accepted.

Reduce Barriers

Despite the characteristics that make artificial intelligence an attractive business tool – from cost-effectiveness and operational efficiency to autonomy – due to its complexity and ambiguities, it is simultaneously difficult to understand and explain, especially to external stakeholders. Therefore, a person who believes they have been harmed by an AI system may be unable to prove that the action or omission related to that system was the cause of the damage incurred. Consequently, such a person is likely to struggle to meet all the requirements prescribed by the legal framework on fault-based civil liability and may ultimately find it entirely impractical or impossible to file a claim for damages.

The aim of the AI Liability Directive is precisely to address this problem by reducing barriers to access to justice when an AI system is the cause of damage. Due to the principle of territoriality, it will apply to all providers of digital services, developers, and users of any AI system operating within the territory of the European Union. The Directive contains two key elements.

Burden of Proof

The first element is that national courts of member states should be authorized to order the disclosure of evidence regarding certain high-risk artificial intelligence systems to the injured party seeking compensation from the service provider or another person subject to supplier obligations. The proposed powers, which would only be available if the claimant can prove that their claim is well-founded and if the claimant has taken all reasonable steps to collect relevant evidence from the defendant, would enable national courts to request that suppliers or users of high-risk AI systems suspected of causing damage present evidence and would stipulate that the court will presume that the defendant, if they do not comply with the court’s order and do not provide evidence in the claim for damages, has not met the principle of due diligence regarding the evidence requested (where this should be a rebuttable presumption if the defendant proves that they exercised due diligence in providing the service).

Due to concerns from organizations regarding the disclosure of their confidential information, particularly trade secrets and proprietary information, it is clear from the proposal text that in determining which information should be disclosed based on a court order, the interests of all parties, including third parties, will be taken into account.

Find Cause and Effect

The second key element of the AI Liability Directive is that national courts of member states would be required to presume that there is a causal relationship between the output of the AI system and the fault of the defendant in the event that the claimant has proven (or the court presumes due to lack of evidence) that the defendant did not exercise the due diligence prescribed by European or national law, aimed at protecting against caused damage; that based on the circumstances of the case, it can be considered likely that the fault influenced the output of the AI system or its lack thereof, and that the claimant has proven that the output or lack of output of the AI system caused damage.

On the other hand, the presumed causal relationship could be rebutted, for example, by providing evidence that the lack of due diligence could not have caused the damage (practically, this further emphasizes the importance of strict record-keeping procedures as outlined in the AI Regulation). Furthermore, the aforementioned presumption of causal relationship should only apply in the case of low-risk AI systems, and only when the court assesses that it is extremely difficult to prove the causal relationship, but not in the case of high-risk AI systems if the defendant can prove that sufficient evidence and expertise are reasonably available to the claimant to establish such a connection.

Include in ‘Defective Product’

Unlike the Proposal for the AI Liability Directive (which is based on fault-based liability), the current European product liability regime is based on no-fault liability to ensure clarity of liability for damage caused by defective products. However, in most cases, the burden of proof for the suffered damage, product defect, and the causal relationship between them still lies with the injured party. The manufacturer generally bears responsibility for product defects, as does the importer for defects in imported products. The Updated Product Liability Directive should now bring AI systems within the scope of the product liability regime in several ways, primarily by including AI systems (as well as goods and services based on AI systems) in the definition of products to which it applies.

If a defective AI system causes damage, compensation is enabled for the injured party without the need to prove the manufacturer’s fault; also, by recognizing that not only hardware manufacturers but also software manufacturers who are providers of digital services can influence the operation of products related to artificial intelligence and that they can also be held responsible, as well as clearly stating that the responsible person can be liable for changes made to products already placed on the market, including when those changes are prompted by software updates or machine learning. It is also essential to establish a regime whereby the responsible person is not exempt from liability if the product becomes defective due to a lack of updates or software upgrades necessary to maintain safety.

The proposed directives, together with the AI Regulation, represent significant regulatory changes for artificial intelligence as they combine the prevention of damage and its compensation. There are many changes, but organizations that take the initiative and adapt in time will be able to continue to create, innovate, and operate with confidence.