European Commission Proposes AI Liability Directive and Modernized Product Liability Directive | Allen & Overy LLP

On September 28, 2022, the European Commission published two proposals to modernize product liability rules in the digital age. The proposal is the first specifically designed to deal with compensation for damage caused by artificial intelligence (AI) systems. General product liability rules are also simplified and extended to address digital products and the circular economy (eg repair, reuse and recycling).

Two directives are proposed:

  • One on the adaptation of non-contractual civil liability rules to AI (AI Accountability Directive); and
  • The other for defective products (Product Liability Directive).

The proposed guidelines are said to complement and complement each other with EU AI law (our article on this can be found here). The Product Liability Directive deals with a producer’s ‘no-fault’ or ‘strict’ liability for defective products (including AI) and the associated compensation for damages. While the AI ​​Liability Directive is about liability arising from “wrongful behavior” of AI systems.

The AI ​​Accountability Directive

The AI Accountability Directive has two new features that aim to address some of the liability challenges of AI:

  • A right of access to evidence. In the case of a traditional product, it can be easy to see what is wrong in the event of a defect. With AI, that may not be the case. To solve this problem, under certain conditions, a supplier of a high-risk AI system (defined in the EU AI law) can be ordered by a court to disclose relevant and necessary evidence about its product. In limited cases, requests may also be made by third parties;
  • A rebuttable presumption of causation. Famous when DeepMind’s AlphaGo defeated a human, some of the moves were ones professional Go players had never considered before. It is an illustration of both the opportunities and challenges posed by AI. As humans, we may not be able to understand the logic behind an AI decision or how one thing led to another. This is one of the reasons why AI liability legislation may have to make a policy decision based on a presumption. Accordingly, under certain conditions and in limited circumstances, national courts would be required to presume, for the purposes of applying liability rules to an action for damages, that the output produced by the AI ​​system (or the inability of the AI ​​system to produce an output) was caused, for example, by the fault of the AI ​​vendor.
  • Who can be sued? There has been a lot of debate about who should be responsible if an AI system fails. These proposals conclude that it should be the providers of the AI ​​systems and, in some cases, the user of the AI ​​systems (each as defined in the EU AI law).

The product liability directive

The revised version Product Liability Directive The proposal adapts existing product liability rules to deal with new types of products and services, such as software or artificial intelligence systems or advanced machinery. Key features include:

  • a right to sue for damages against the manufacturer of a defective product that caused death, personal injury (including medically recognized psychological harm), property damage, or loss or corruption of data;
  • software or AI systems and AI-enabled goods are explicitly referred to as “products” under this proposal, and injured individuals can seek compensation if the software or AI causes damage ;
  • in cases where the court finds that a person bringing an action has difficulty in proving a defect or a causal link between defect and damage due to technical or scientific complexity, the burden of proof will be reduced (although the person prosecuted can challenge it). This proposal to shift the burden of proof aligns the proposed Product Liability Directive with the AI ​​Liability Directive;
  • not only hardware manufacturers, but also software providers and digital service providers that affect the operation of a product (for example, the provider of a navigation service in an autonomous vehicle) may be liable managers;
  • manufacturers can be held liable for changes they make to products they have already released (for example when those changes are triggered by software updates or machine learning). A person modifying a product that has been placed on the market or put into service may, in certain cases, be considered as the manufacturer and therefore liable for damages;
  • claim thresholds and caps on compensation levels are removed. According to the rules in force, to be considered for a claim, the damage caused by the defective product must be at least 500 euros, and Member States may impose a limit on the producer’s total liability for damage resulting from death or bodily injury (this amount cannot be less than 70 million euros);
  • other parties such as importers, authorized representatives, fulfillment service providers or distributors may be liable for defective products manufactured outside the EU or where a manufacturer cannot be identified.

The press release is available here. The AI ​​Accountability Directive is available here and the Q&A on it here. The Product Liability Directive is available here and the questions and answers about it here.