Spotlight: The proposed Product Liability Directive and AI Liability Directive

January 13, 2023

Share on LinkedIn Share on Facebook Share on X

By guest author: Stefan Lohn

Foreword by Chris Occleshaw, international product recall consultant at Sedgwick:

Welcome to the latest edition of Spotlight. Spotlight is our way of sharing insights and perspectives from our strategic partners – lawyers, insurers, risk managers and crisis communications experts across industries – on issues that have the potential to influence a company’s view on in-market incidents and crisis management. In this edition, we are joined by Dr. Stefan Lohn, Counsel at Clifford Chance, who shares his thoughts on the European Commission’s proposals to revise the framework for product liability and artificial intelligence (AI) civil liability.

What is the purpose of the proposed Product Liability Directive and AI Liability Directive?

The European Commission adopted the proposed Product Liability Directive (P-PLD) and proposed AI Liability Directive (P-AILD) on 28 September 2022. The proposals were released as a package to update civil liability rules for the digital age. Together, the documents cover specific challenges that artificial intelligence (AI) systems pose and different aspects of non-contractual liability. Moreover, they expand the existing product and AI civil liability framework to reflect advancements in information technology.

If adopted as expected by the European Parliament, the proposed directives will require businesses and enterprises to reassess their liability risks and how they manage documentation and information, particularly with regard to disclosure obligations.

What are the main features of the P-PLD?

The P-PLD governs non-contractual civil liability for defective products. The revised P-PLD maintains several important features of the existing Product Liability Directive (PLD). However, it also addresses shortcomings created by evolving definitions and concepts of products since the earlier PLD was introduced in 1985 and updates them where necessary.

The proposal notes that the framework of the existing PLD may no longer be suitable for some products in the modern digital economy and circular economy, including software, smart devices and autonomous vehicles.

The P-PLD changes the terminology in the current PLD from “producer” to “manufacturer.” It also assigns strict liability to manufacturers and importers for damages caused by a product or its component. However, the P-PLD aligns its terminology with the existing framework of the Decision 768/2008/EC regarding the marketing of products.

Familiar elements such as the possible liability of multiple economic actors continue to be a part of the P-PLD. However, the group of potentially liable economic actors is extended to include service providers that warehouse, package, address and dispatch products if at least two of these services are rendered. This may give rise to a potential liability for distributors, online platforms offering a product and fulfilment service providers who may believe they are exempt from the existing PLD.

Moreover, intangibles, such as software integrated in a product, are deemed to be a component of that product. That means software developers are subject to the same potential liability as conventional product manufacturers. Hence, software providers and providers of digital services, such as satellite navigation service in an autonomous vehicle, are also subject to the new PLD.

In addition, AI systems and AI-enabled goods fall within the scope of the P-PLD, thus enabling an injured person to obtain compensation without having to prove fault if a defective AI system or enabled product causes damage.

The P-PLD aims to further alleviate the burden of proof for injured persons in complex matters, including cases involving AI systems or when products fail to comply with safety requirements.

The P-PLD also expands the definition of damage to include loss and corruption of data. Furthermore, it asserts that manufacturers will be liable for changes to products already placed on the market, including, but not limited to, changes by means of software updates or machine learning.

Finally, the P-PLD disposes of prior limitations on the minimum and maximum amount recoverable. The general limitation term remains unchanged, but the proposal extends the absolute time limit for personal injury claims under the PLD from 10 to 15 years if an injured person had a latent personal injury and could not assert a claim in the 10-year period.

What are the main features of the P-AILD?

The P-AILD concerns non-contractual damage claims against providers of AI systems and aims to lay down uniform requirements for damage caused by the use of AI systems. In this context, an AI system includes software that employs machine learning, inductive programming and statistical approaches.

The P-AILD is intended to apply to non-contractual civil law claims for damages caused by an AI system, including claims related to a fault-based civil liability process. According to the proposal, a harmed person seeking recourse for a damage caused by an AI system shall be furnished with effective substantive and procedural means to identify potentially liable persons and respective evidence relevant to the respective claim.

There are two main components of the P-AILD. First, the P-AILD enables courts to order the disclosure of relevant evidence about the AI system that is suspected of having caused the damage. The disclosure obligation is addressed to the provider of an AI system as well as a person who is subject to the provider’s obligations as set out the AI Act or a user pursuant to the AI Act.

The disclosure obligation is complemented by procedural means to preserve evidence, but also includes provisions to safeguard the proportionality of disclosing the evidence. This framework ensures a balance between the justified interests of the party seeking disclosure with the interests of the disclosing party in its trade secrets or otherwise confidential information. Finally, the P-AILD introduces a presumption of a violation of a duty of care if the defendant fails to comply with a disclosure obligation, though the defendant can rebut this assertion.

The second component is the establishment of a presumption of causation between the fault of the defendant and the output produced by the AI system or the failure of the AI system to produce such output. The claimant must be able to meet three criteria: 1) demonstrate the fault of the defendant, 2) show the fault is reasonably likely to have been influenced by the output of the AI system and 3) demonstrate a causal link between the output, or non-output, of the AI system and the damage.

This presumption of causality is designed to ensure a level of protection for an injured person similar to situations where AI is not involved. The defendants are given the opportunity to rebut these allegations of causation.

What’s next?

The two proposals from the Commission now must pass through the European Parliament and the Council. Once the P-PLD and the P-AILD have been adopted, member states will then need to follow through with adoption of the directives.

For the P-PLD, the Commission proposes a maximum period of 12 months for the implementation by member states. A maximum period of two years has been proposed for the P-AILD.

Download a copy of this Spotlight feature.

About our guest author, Dr. Stefan Lohn

Dr. Lohn specialises in commercial arbitration and litigation and advises on supply chain compliance and risk. His practice focuses on international and domestic warranty and product liability matters including automotive, capital goods and electronics, as well as commercial contracts.

Dr. Lohn is a member of the German Institute of Arbitration (DIS), as well as Clifford Chance’s Automotive, Capital Goods, and Product Liability sector focus groups. He publishes frequently on supply chain compliance and risk as well as product liability issues.