Is product liability law fit for AI?

Is product liability law fit for AI?

My previous article considered professional negligence and who should be liable when AI causes harm. Product liability raises a related question: can claimants recover compensation for harm caused by defective AI without having to prove negligence?

Under the Consumer Protection Act 1987 (CPA), claimants can recover compensation for damage caused by defective products without proving fault. However, the legislation was designed around physical products, creating uncertainty about its application to AI.

The current law

The UK Jurisdiction Taskforce (UKJT) illustrates the problem through an AI-powered fridge. When a fridge incorrectly identifies spoiled oysters as safe to eat, causing serious injury, the manufacturer and developer of its bespoke AI component may be liable under the CPA. The provider of the underlying foundation model, however, is unlikely to be liable: it supplied the model as a cloud service, rather than a product or product component.

The problem is that liability may depend on how AI is supplied, rather than where the fault lies.

More broadly, the CPA's application to standalone software remains uncertain. Where AI is embedded in a physical product, such as a medical device or vehicle, the Act may apply. Software supplied digitally presents a less straightforward case, potentially leaving claimants without access to strict liability.

The scope of recoverable damage is also limited. The CPA covers death, personal injury and certain damage to privately used property, but excludes pure economic loss and damage to property intended mainly for business use. This limits its usefulness where AI corrupts data or causes financial loss.

Finally, AI supply chains complicate the identification of liable parties. Responsibility may be distributed between developers, integrators and deployers, while the CPA only imposes liability on specified categories of defendant.

Proposed reforms

The Law Commission proposes extending the definition of a product to include software and AI systems, whether embedded in physical products or supplied digitally. This could address the distinction illustrated by the fridge example, bringing standalone AI within the framework.

The proposals would also expand the categories of potentially liable parties to include fulfilment service providers, online marketplaces and UK authorised representatives, alongside a proposed fallback category of “core participants”. They would clarify liability for products substantially modified after supply.

To address the difficulty of proving defectiveness, the Commission proposes a clearer assessment framework and a rebuttable presumption of defect in certain cases involving exceptional evidential difficulties.

What does this mean for lawyers now?

Practically for lawyers, then, identifying how an AI system was supplied, who developed and integrated it, and what damage occurred will be central to assessing whether a CPA claim is available. Where it is not, alternative claims in negligence or contract may need to be considered.

For firms developing or deploying AI, the proposals reinforce the importance of reviewing supplier contracts, documenting modifications and retaining testing records. These measures can help manage existing risks and prepare for potential reform.

A framework under pressure

Product liability law provides a route to compensation without proof of negligence, but its focus on physical products may leave gaps when AI causes harm.

The Law Commission's proposals seek to address those gaps. The central question is whether the law can provide effective redress when liability depends not only on what caused the harm, but on how the technology was supplied.

Megan is a Contributing Writer at Best Practice. She is currently a second-year Law student at Queen’s College, University of Oxford. Alongside her studies, she is an Associate Editor at the OUULJ, and is pursuing a career at the Bar. Her areas of legal focus are in tort, copyright and public law, particularly through the lens of transformative AI developments.