Anticipating the revised EU Product Liability Directive and the new EU Artificial Intelligence Liability Directive

It has been a busy few years for legislative developments in life sciences in the EU. We have seen the introduction of landmark new pieces of legislation, such as the Medical Device Regulation,[1] the In Vitro Diagnostic Medical Device Regulation,[2] and the Clinical Trials Regulation.[3]


First published in our Biotech Review of the Year publication (Issue 10).

We have had proposals for cutting edge new legislation, such as the Artificial Intelligence Regulation and the European Health Data Space Regulation, which break new ground in terms of the technologies they address and the problems they aim to solve. A consultation has also been conducted regarding major revisions to the general pharmaceutical law legislation, which has yet to result in a legislative proposal. The Orphan and Paediatric Regulations are also under review.

To add to that list, the European Commission has announced two significant new legislative proposals. These are the proposal for a revised Product Liability Directive (Revised PLD) and the proposal for an Artificial Intelligence Liability Directive (AILD).

The Revised PLD and the AILD will not be blockbuster pieces of legislation along the lines of the GDPR or the Medical Device Regulation. They do not establish lengthy regulatory codes for companies to get to grips with. However, the Revised PLD will nonetheless introduce new rules on liability which will impact economic operators across product supply chains, and both the Revised PLD and the AILD will make some noteworthy changes to litigation procedure in EU Member States. Life sciences companies ought to be aware of these changes – especially those with products which incorporate software or AI as a significant component.

Revised PLD

The Revised PLD will replace the existing Product Liability Directive,[4] in existence since 1985. The Revised PLD preserves the basic principle of strict liability for defective products which cause harm to consumers, which is the core of the 1985 Product Liability Directive, but it makes a number of key changes which are intended to bring product liability “into the digital age.”

The following are some highlights:

  • Standalone software and AI systems will be expressly considered “products” for the first time, allowing claimants to obtain compensation under the Revised PLD if such systems are defective and cause them harm;
  • “Harm” for which a claimant can obtain compensation will include medically recognised psychological harm and the loss or corruption of electronic data (where it is not used exclusively for professional purposes);
  • A product will be considered “defective” if the manufacturer has not designed and maintained it to a sufficiently high cybersecurity standard. A product incorporating software will also become “defective” if the manufacturer fails to release necessary cybersecurity updates;
  • The list of entities which could be liable for defective products under the Revised PLD will expand significantly. In addition to manufacturers of defective products, importers and distributors can sometimes be liable for defective products under the 1985 Product Liability Directive. The Revised PLD will also place authorised representatives,[5] fulfilment service providers (FSPs) and even online platforms which facilitate transactions between traders and consumers in the firing line in certain circumstances.

As well as increasing the scope of product liability by expanding the definitions of “product”, “harm” and “defective”, and adding to the list of potential defendants to a defective product liability claim, the Revised PLD will also make changes to procedural law to make it easier for consumers to bring claims in certain circumstances.

Under the 1985 Product Liability Directive, consumers injured by a defective product must bring a claim within a limitation period of 10 years from the product being put into circulation. Under the Revised PLD, if consumers are unable to commence proceedings within 10 years because of the latency of the personal injury caused by the defective product, the limitation period will be extended to 15 years. It can be quite common for personal injuries caused by defective life sciences products to be latent – such as where an implantable medical device gradually leaches a hazardous substance into the body or where a medicinal product has a long term adverse effect on internal organs such as the liver – so life sciences companies should regard this change as a potentially significant increase in their product liability exposure.

As well as extending limitation periods, EU Member State courts will have the power to order disclosure of information in the context of defective product liability litigation. To litigators with experience before the English courts, this will not sound like a radical development, but comprehensive disclosure of relevant documentary evidence is not a routine part of litigation in many EU Member States. Generally, parties are only obliged to disclose the evidence they intend to rely on, and parties can then apply to the court for an order requiring the other party to disclosure further specific documents. The Revised PLD is phrased such that claimants in defective product litigation may be able to obtain disclosure of all “relevant” evidence relating to the litigation, which is much closer to the English procedural approach in civil litigation.

In order to remedy perceived expertise and resource imbalances between consumer claimants and economic operator defendants, the Revised PLD will also shift the burden of proof onto the economic operator defendant in certain circumstances. Specifically, a product will be presumed to be defective where:

  • The defendant has failed to comply with an order for disclosure made under the Revised PLD;
  • The defendant has failed to comply with requirements under mandatory product safety laws which are intended to protect against the kind of harm which the claimant alleges they have suffered; or
  • The allegedly defective product has obviously malfunctioned under normal conditions of use.

It will also be rebuttably presumed that the product was defective, and/or that its defectiveness caused the harm to the claimant, where the court considers that it would be excessively difficult for the claimant to prove either of these assertions. First the claimant will have to succeed in showing that: (a) the product contributed to the harm; and (b) it is likely that the product is defective and/or that its defectiveness caused the harm in question.


Even more so than the Revised PLD, the AILD is brief. Its core operative parts cover scarcely more than two sides of A4 paper. However, it will introduce some quite striking changes to procedural law in EU Member States in relation to non-contractual fault-based claims concerning damage caused by AI systems.

That phrase “fault-based claims” deserves exploration. The Revised PLD (and the existing 1985 Product Liability Directive) create liability in relation to defective products which is not dependent on the defendant being at fault. The AILD will not apply to such claims. The AILD will create procedural rights for claimants in the context of claims where it is asserted that the defendant was at fault in the use or production of an AI system – either deliberately or negligently causing harm.

The purpose of the AILD is to make it easier to bring fault-based claims in respect of harm caused by AI systems, where the complexity and opacity of such systems would make it difficult for claimants to discharge the evidential burden. In effect, the AILD will confer two new procedural rights on claimants harmed by the use of AI systems. Interestingly, where the Revised PLD prescribes the kind of harm which is recoverable under it, the AILD does not. Seemingly, claimants can rely on the rights created by the AILD in relation to non-contractual fault-based claims for any kind of harm, including damage to business property and perhaps even pure economic loss caused by relying on the recommendations of an AI system.

The first procedural right created by the AILD is the right to seek pre-action and/or third party disclosure of relevant evidence from the “providers” and/or the “users” of “high risk” AI systems which are said to have caused harm to the claimant. This can include disclosure of the logs automatically generated by the AI system which will have to be kept under the proposed Artificial Intelligence Regulation. The terms “provider”, “user” and “high risk” will also have the meaning conferred on them by the proposed Artificial Intelligence Regulation.

If the “provider” or “user” of a high risk AI system is the defendant in a fault-based action and fails to comply with a disclosure order made under the AILD, there will be a rebuttable presumption that the defendant did not comply with its duty of care to the claimant.

The second procedural right created by the AILD echoes aspects of the Revised PLD: a reversal of the burden of proof consisting of a rebuttable presumption that the defendant’s (actual or presumed) failure to comply with its duty of care to the claimant caused the harm which the claimant alleges. This rebuttable presumption will arise where the following criteria are met:

  • The claimant has established, or it has been presumed, that the defendant did not comply with a duty of care to the claimant which is directly intended to protect against the harm which has occurred;
  • It is reasonably likely in the circumstances that the non-compliance has influenced the output of the AI system or its failure to produce an output; and
  • The output of the AI system, or its failure to produce an output, has caused harm to the claimant.

Where the AI system which is said to have caused harm is a “high risk” AI system within the meaning of the Artificial Intelligence Regulation, non-compliance with the requirements of the Artificial Intelligence Regulation will be considered a failure to comply with a “duty of care” to the claimant for the purposes of the presumption.

Where the AI system which is said to have caused harm is not a “high risk” AI system within the meaning of the Artificial Intelligence Regulation, then the presumption will only apply where the national court considers that it would be “excessively difficult” for the claimant to prove that the defendant’s wrongful acts or omissions were the cause of the harm suffered.


Evidently, the EU has lost none of its enthusiasm for piling on multitudinous new legislation in response to changes in the European economy. There is a lot coming down the pipeline for life sciences companies to stay on top of. For the time being, the Revised PLD and the AILD are only legislative proposals, and they may undergo further changes.

The Revised PLD is an evolution rather than a revolution in the field of product liability. Nonetheless, it will bring some changes of which life sciences companies should be aware. Life sciences companies should be aware that they will be exposed to potential product liability claims for a significantly longer period of time where the damage caused by a defective product is latent. Life sciences companies should also consider their own potential liability as distributors of original equipment manufacturer (OEM) products. For instance, pharma companies which distribute OEM software medical devices intended for use with their medicinal products should be aware of their potential liability.

Where the Revised PLD is evolutionary rather than revolutionary, the AILD is genuinely new, but in many ways it is simply a liability-related adjunct to the much more ambitious Artificial Intelligence Regulation (although narrow sections of the AILD do apply to AI systems outside the scope of the Artificial Intelligence Regulation). Much more would need saying to fully explore the proposed scope of the Artificial Intelligence Regulation, but a brief internet search or a perusal of previous Bristows publications will explain quite how broad that scope is. Any business which will be affected by the Artificial Intelligence Regulation should have the AILD on its radar. In particular, businesses should be prepared for applications for disclosure under the AILD, even where they are not a party to the litigation. They should also be aware that non-compliance with the Artificial Intelligence Regulation will increase their potential exposure to non-contractual fault-based liability claims, due to the rebuttable presumption mechanism created by the AILD.


[1] Regulation (EU) 2017/745 on medical devices

[2] Regulation (EU) 2017/746 on in vitro diagnostic medical devices

[3] Regulation (EU) 536/2014 on clinical trials on medicinal products for human use

[4] Council Directive 85/374/EEC on the approximation of the laws, regulations and administrative provisions of the Member States concerning liability for defective products

[5] Article 11(5) Medical Device Regulation already makes authorised representatives liable for defective medical devices, but it has never been clear how exactly this provision is intended to work