AI as a product: an analysis compares China, the EU and the UK

In a commentary on the Oxford Business Law Blog, published on Oxford Law Blogs, Mohammad Samiul Huq compares Chinese and EU approaches to AI product liability and the questions facing the UK. His account of the Chinese guidance draws on an analysis by the European Commission’s IP Helpdesk and, for some references, China Law Translate’s unofficial translation. The primary texts discussed have not been independently examined.

According to Huq, guidance issued by China’s Supreme People’s Court in September 2026 keeps no-fault product liability tied to AI incorporated into physical goods, such as robots or autonomous vehicles. For standalone generative AI services, the starting point remains fault-based liability, ordinarily requiring proof of fault, damage and causation. The account identifies autonomous learning, updates, user control, safety standards and warnings as factors in assessing defects in physical products. In autonomous or assisted-driving disputes, a party controlling relevant driving data may be required to provide complete records.

For the EU, Huq describes the adopted Directive (EU) 2024/2853, which includes software within the definition of a product, including when accessed through the cloud. The framework considers subsequent learning and continued manufacturer control, while providing evidence disclosure mechanisms and rebuttable presumptions; the commentary does not specify its application timetable. In the UK, the Law Commission is reviewing the regime under the Consumer Protection Act 1987. For the author, reform choices concern bringing standalone AI within no-fault liability, defects arising after supply, control over updates and difficulties of proof. These are questions for reform, rather than UK amendments already enacted.