TL;DR

The Law Commission opened a consultation on 8 October proposing that software and AI systems count as “products” under the Consumer Protection Act 1987 (Part 1), the regime intended to help those hurt by faulty goods get compensation. Social media design features, recommender algorithms included, would also come within scope, and online marketplaces would face liability. Responses close on 14 January 2027, with final recommendations due in late 2027.

The 1987 law predates smartphones, apps and AI, and the Commission says it was built for physical goods. Under the proposals, an AI system would be treated as a product whether it is built into hardware, downloaded or reached over the internet.

What would change for AI suppliers

In our view, four of the Commission’s six key proposals matter most for businesses that build or sell AI:

• A wider product definition, covering software, AI systems and the design elements of online platforms, however they are delivered.
• A presumption of defectiveness in technically complex cases where a claimant has exceptional trouble getting evidence. A defendant could try to rebut it, or accept it, in which case a court would award part of the damages through a new way of sharing the risk.
• Clearer rules on psychiatric harm, both for people directly injured and for those who become ill after seeing a loved one harmed or killed.
• Guiding principles for courts on judging defectiveness, plus a refreshed set of factors for them to weigh, which the Commission hopes will cut litigation costs.

Producers keep their defences, adapted for digital products. That includes the “development risks” defence, which protects a producer where a defect was undiscoverable at the point of supply, or at the point it later developed.

Professor Solène Rowan, who leads the Commission’s commercial and common law work, said the law “must evolve” with technology, and that the aim is compensation for people harmed alongside “greater clarity and certainty for businesses”.

How it fits with existing thinking

It reads differently from the UK Jurisdiction Taskforce’s July legal statement on AI harms. That statement said negligence and professional liability principles can apply where AI causes harm, but listed as unresolved whether product liability law covers standalone AI software. In our view, moving AI into product liability would shift attention from who was careless to whether the system itself was defective, which puts more weight on how it was designed and tested.

Looking forward

In our view, the presumption of defectiveness is the proposal UK AI developers should study first, because the Commission aims it at technically complex claims where evidence is hard to obtain, a description that fits many disputes over AI. Firms that deploy third-party models should also check where their contracts leave them if software and AI systems become products in law. The consultation is open to businesses, consumer groups and lawyers until 14 January.