UKJT Legal Statement on liability for AI harms: Key takeaways

As organisations increasingly rely on AI systems to generate content, deliver advice, automate processes and inform decisions, a fundamental legal question arises: who bears responsibility when AI causes harm? While policymakers around the world continue to debate AI-specific regulation, disputes arising from AI deployment are likely to reach the courts long before any comprehensive liability regime is introduced.

Following a public consultation, in July 2026 the UK Jurisdiction Taskforce published its Legal Statement on liability for non-deliberate AI harms (“Statement”).

The Statement is not legally binding but provides guidance on how the law of England and Wales can, in the absence of a specific regime, deal with cases relating to non-deliberate harm caused by the use of AI. Its central conclusion is a reassuring one for litigators: English law is generally sufficiently flexible to deal with AI-related disputes through the application of familiar legal doctrines including contract, negligence, professional liability, and causation. The Statement does not cover economic torts, competition claims, IP rights, data protection or issues that arise out of the use of AI by public authorities.

We summarise some of the key points below:

Basis of liability 

1. Contract: Parties can voluntarily assume responsibility for AI harms by contract; or 

2. Negligence: Absent a contract, non-deliberate loss resulting from the use of AI will likely be determined by the ordinary principles of negligence, notwithstanding the AI context. A claimant will need to show that the defendant owed them a duty of care, that the defendant breached its duty, and that the breach caused the claimant to suffer loss. The existence of a duty of care is fact sensitive, and the Statement acknowledges that, in practice, negligence claims concerning the use of AI will generally involve professional negligence or statements made by AI. 

Vicarious liability 

Because an AI system is not a legal person, no one can be vicariously liable for its actions or failures. However, the Statement outlines that an employer could be held vicariously liable where harm occurs as a result of an employee’s wrongful use of AI.

Professional liability 

The obligation on professionals, such as lawyers, architects and doctors, to exercise reasonable care and skill in carrying out their duties applies to their use of AI. The Statement identifies that a professional could be negligent for: 

1. The inappropriate use of AI, such as using an unsuitable model, failing to conduct proper due diligence, or for failing to test AI or validate outputs; also 
2. Failing to use AI where a competent member of their profession would have done so.

Strict liability 

In the case of strict liability for death, personal injury or damage to property caused by a defective product, strict liability would only arise where AI is incorporated into a physical product. 

Causation

Where a party has acted in breach of contract or negligently, the breach/negligence must have caused the other party’s loss. 

Whilst AI’s opacity means there may be issues from an evidential perspective (i.e. the reasons why AI has produced a certain outcome may not be discoverable), the “but for” test remains the determinative test for establishing factual causation. This requires a claimant to prove that “but for” the breach/negligence, the loss would not have occurred.

The Statement says that, generally speaking, AI developers are unlikely to be liable for loss flowing from its use by third party bad actors unless the “AI in question was obviously dangerous”, or “the person had the power to prevent the misuse but failed to do so when it should have been done”.  

The availability of contributory negligence will be fact specific. Whilst commercial users are more likely to be found to be contributory negligent than non-commercial users, contributory negligence would be appropriate where a user relies on AI output “where it is obvious that the consequences of an AI error will be serious. For example, if an AI system suggested a driving route that involved crossing soft, tidal sands, and a non-commercial User – knowing of the likely hazard – crossed the beach anyway, a finding of contributory negligence would seem appropriate in response to a claim for the cost of recovering their stuck vehicle ”.

False statements/AI Chatbots 

A person who holds out an AI as communicating on their behalf could be liable for false statements made by the AI. Liability could also arise where a person is deemed to be a publisher of defamatory statements generated by AI. 

What does this mean?

For commercial parties, the practical message is clear. Contractual allocation of risk will remain the primary mechanism for managing liability within AI supply chains, while negligence principles will continue to provide the principal route to recovery where contractual remedies are unavailable. Businesses adopting AI technologies should focus on governance, record-keeping, testing, oversight and clear contractual arrangements, all of which are likely to assume increasing significance in any future dispute.

For practitioners, the report provides affirmation that English law already possesses many of the tools needed to address AI harms. As AI-related disputes become more common, the challenge for courts will be less about inventing new principles and more about applying established ones to technologies whose capabilities, autonomy and impact continue to evolve at pace.