Megan Higgins and David McIlwaine of Pinsent Masons made the comments after the UK Jurisdiction Taskforce (UKJT) published a draft legal statement on liability for AI harm.
The document is open for consultation until February 13, 2026 and is not binding in court. The UKJT aims to provide some clarity about how emerging technologies interact with current legislation. No views are expressed on areas where there is room for reform. In this case, the opinion is designed to help you understand the circumstances in which a business may be liable for, or may be able to bring a claim against others, for AI-related harm under private law in England and Wales.
This statement is premised on the basis that AI systems do not themselves have legal personality under UK law and therefore cannot be held legally responsible for physical or economic harm. However, according to the UKJT, individuals and businesses can be held liable for harm related to the use of AI in a variety of circumstances.
Meghan Higgins of Pinsent Masons said: “One of the most interesting aspects of this paper is its discussion of how the use of AI can create uncertainty in litigation situations.”
“The authors explain that current liability systems typically impose legal liability on individuals or businesses that cause harm through a voluntary act or omission. AI systems are autonomous, meaning their outputs are not predetermined or programmed, and these systems “The use of systems can have consequences that were not anticipated by the parties involved in the development, training, and implementation of the system. This can make it difficult to determine which party is responsible for any damage caused, and to what extent,” she said.
“The opacity of AI systems, that is, the difficulty of understanding and proving how a result is reached, can create further difficulties in assessing causation and foreseeability in the context of litigation. This will be particularly acute in complex supply chains where AI systems may be developed and trained by multiple parties,” Higgins added.
According to UKJT, issues of liability often depend on the terms of the contract between the parties. “The extent of liability and the ability to pass losses upstream is typically determined by warranties, indemnities, limitations and exclusions,” the report said.
The UKJT said contracts are particularly important when determining liability for economic damage, as the law provides that liability for economic damage does not arise “unless there is a ‘special relationship’ between the parties, one of whom is voluntarily liable to the other”, and said this “generally (but not always) involves a contract”.
The UKJT said the potential for physical harm from the use of AI is “most evident in the context of ’embodied AI’, such as self-driving cars, medical robots and assembly line machines”, where the output of the AI is “directly controlled”. However, the report said physical harm could also occur in other situations, such as when the use of AI in cancer detection results in “false positives or false negatives.”
Even where contracts do not specify liability, companies and people may be subject to extra-contractual obligations to protect against AI-related harm, the task force said.
In most cases, liability only arises when negligence is involved. The UKJT stated that those involved in developing, deploying and operating AI systems are not liable for damage caused by the system in the absence of negligence on their part. However, there is an exception to the UK’s product liability regulations under the Consumer Protection Act 1987, which impose “no-fault” liability for physical damage resulting from a defective product. It is currently unclear how these rules will apply to IT systems, but the Law Commission is consulting on potential reforms to the system.
The UKJT said that an employee’s careless use of AI could make the employer liable under liability legislation for damages arising from its use, even if the employer itself was not at fault.
It said AI chatbots could also be held liable for false statements made under the theories of negligent misrepresentation and defamation.
The success of claims for liability for damages caused by AI will depend on whether it can be proven that those damages were caused by the use of AI. However, the UKJT said the autonomous nature of AI and its opaqueness could make it difficult to understand exactly why a particular outcome occurred, and expert evidence may need to be provided to assess whether there is a causal relationship.
David McIlwaine of Pinsent Masons said: “The UKJT statement provides a very useful overview of how the use of AI can give rise to liability in different circumstances and how courts may allocate liability.”
“The rapid proliferation of AI in many sectors means that its use is likely to become firmly embedded in business practice before the courts have had a chance to consider these issues. UKJT remains optimistic that UK private law principles are sufficiently flexible that the courts will be able to address uncertainties as they arise,” he said.
“To the extent that AI raises special questions regarding causation, evidentiary presumptions may develop and the need for experts who can help judges understand the particular AI system at issue may increase. Companies should recognize that the use of AI may pose unique risks in litigation situations, but the easiest way to address them is at the beginning of the relationship between the parties to the contract, particularly warranty, liability and exclusion clauses,” McIlwain added.
