On February 10, 2023, the European Parliament published its briefings of Legislation in Progress on new Product and AI liability directives. They allow us to learn more about the approaches of the Union on such matters.
Image by vectorjuice on Freepik
First, what are these two directives and their differences?
Two draft directives are complementary for each other. The European Commission proposed them on September 28, 2022. (1) They wish to adapt the existing liability rules – Product Liability Directive (the PLD) in 1985 as the harmonised fragmented legal protection on damage caused by defective products to new digital technologies. The proposed directive on liability for defective products shall revise the PLD to modernise the existing EU no-fault-based (strict) product liability regime. It applies to claims made by private individuals against the manufacturer for damage of defective products (2) The AI liability directive may ensure broader protection for damage caused by AI systems. Then, it will alleviate the burden of proof in compensation claims pursued under the national fault-based liability regimes. They apply to claims made by a natural or legal person against any person, for fault influencing the AI systems causing the damage.
Second, the new Product Liability Directive (the PLD)
For the definition of products and economic operators
The new PLD widens the definition of products provided in article 4(1). As a result, it shall cover the software including AI systems but free open-source software for non-commercial purposes and source code of software, digital manufacturing files, digital services.
Furthermore, it also broadened the scope of liable parties in articles 4(16), 7. It provided a layered approach to liability subject to the qualification of “economic operators”. Particularly, the manufacturer, including any economic operator who has substantially modified the product outside the control of the main manufacturer, should be liable for the defects in their product or components. If the manufacturer is established outside the EU, the importer and the authorised representative in the EU shall be liable for the defect. Moreover, if the latters are established outside the EU, the fulfilment service provider shall be the last resort. In addition, the claimant can hold the distributor liable for a defective product if he/she fails to identify any other of above operators. Also, the online platforms may be liable if they perform the role as manufacturer, importer or distributor.
For the definition of damage
The revised PLD would expand the definition of damage (article 4(6)), by including material losses, resulting from death or personal injury, including psychological health with the extended liability period up to 15 years, property damage, and loss or corruption of data. The extent of liability is quite large. The developers would be responsible for emerging technologies, learning independently, for the deployment update or lack thereof. Defectiveness is presumed when (1) a manufacturer fails to comply with obligation of information, (2) a product does not comply with mandatory safety requirements, (3) obvious product malfunction causes damage. The claimant can presume the causality when damage is typically consistent with the defect in question, or technical and scientific complexity causes excessive difficulty in proving liability, e.g. “black box” of AI systems.
For the liability exemption
The economic operators can be exempted from the liability, for which they carry the burden of proof, when (1) they did not put the product into circulation, (2) the defect did not exist when they placed the product on the market; or (3) the state of technical knowledge at the time of placing the product on the market made it impossible to discover the defect. The exemptions shall not apply to the defects within the manufacturer’s control linked to a related service, software including software updates or upgrades, or a lack of updates or upgrades necessary to maintain safety.
Sources: https://www.europarl.europa.eu/thinktank/fr/document/EPRS_BRI(2023)739341;
https://www.europarl.europa.eu/RegData/etudes/BRIE/2023/739341/EPRS_BRI(2023)739341_EN.pdf
Third, the AI liability directive
The proposal seeks to harmonise non-contractual (or “extra-contractual”) civil liability rules for damage caused by AI systems, irrespective of whether they are defined as high-risk or not. It provides compensation regardless of a contractual link between the victim and the liable person. The directive would create a presumption of causality giving claimants seeking compensation for damage caused by AI-systems a more reasonable burden of proof, under certain cumulative conditions:
- The claimant has demonstrated that the non-compliance with a certain EU or national obligation relevant to the harm of an AI system caused the damage (article 4(1)(a));
- It must be reasonably likely that, on a case by case basis, the defendant’s negligent conduct has influenced the output produced by the AI system or the AI system’s inability to produce an output that gave rise to the relevant damage (Article 1(b)).
- the claimant has demonstrated that the output produced by the AI system or the AI system’s inability to produce an output gave rise to the damage (Article 1(c)).
Furthermore, to remedy this, the AI liability directive would give national courts the power to order disclosure of evidence about high-risk AI systems that are suspected of having caused damage (Article 3(1)).
Sources:
https://www.europarl.europa.eu/thinktank/fr/document/EPRS_BRI(2023)739342
https://www.europarl.europa.eu/RegData/etudes/BRIE/2023/739342/EPRS_BRI(2023)739342_EN.pdf
For more information about the AI systems on AstraIA Gear: