Who pays when AI gets it wrong?

Answer

Summary: The UK Jurisdiction Taskforce says English law is broadly capable of dealing with AI-related harms without new legislation. It suggests that contracts will often decide who bears the risk when AI causes loss, while negligence law will fill the gaps where no contract exists. The statement also explores issues such as professional liability, causation and false AI-generated statements, providing a framework for resolving disputes as AI becomes more widely used.

Ask clients and colleagues what topic is most important to them right now, and the answer is likely to be AI – how to use it, what tool to buy and what it means for the future. Meanwhile, the focus of the English courts has been predominantly on how the use of AI tools impacts the legal process. Judgments have focused on topics such as AI hallucinations, the risks of using AI to draft witness statements or prepare for giving evidence, and the potential loss of privilege and confidentiality when using public AI tools. None of this speaks to the much bigger question of the substantive claims AI is likely to generate and how they might ultimately be decided. It’ll take time for such claims to work their way through the courts. There’s also no AI-specific liability regime in England.

The UK Jurisdiction Taskforce (UKJT) has stepped into the breach.

The UKJT is an industry-led initiative established by the LawtechUK panel in 2020 to promote the use of English law and jurisdiction for technology and digital innovation. It issues statements to explain how the common law is likely to deal with private law problems emerging from new technology. In July, it published a legal statement on liability for AI harms. The statement follows a public consultation held earlier in 2025, to which Herbert Smith Freehills Kramer LLP (HSF Kramer) contributed, and addresses a question of growing commercial importance: in what circumstances, and on what legal basis, will English law impose liability for loss resulting from the use of AI?

The focus of the statement is on the application of the law of negligence to physical and economic harms caused by AI. Its central message is that English law is already flexible enough to handle the challenges posed by AI – not least because the common law has a long track record of adapting to novel and disruptive technological change. Its conclusion, at a high level, is that the law is generally well-equipped to deal with the liability questions that AI is likely to raise. That said, the statement notes that the existing framework has its limitations.

Adaptability is not the same as legal certainty. In practice, those affected by AI-related harms may find themselves relying on long-established principles to resolve disputes that are framed in entirely novel ways.

Key areas addressed by the UKJT and its position

The statement focuses on whether and in what circumstances those who haven’t set out deliberately to cause harm, such as developers, businesses and professionals, may nonetheless be liable when AI causes loss. It addresses several core areas of law.

Contract and negligence

Contract is likely to be the primary basis on which liability is allocated between parties in an AI supply chain. Where a contract exists, the question of who bears the risk of AI-related harm will depend on the terms agreed. For businesses operating in AI supply chains, a contract will often be the key point of reference. Businesses that invest in careful contractual drafting now – covering warranties, indemnities, liability caps and AI-specific risks – are likely to be in a much stronger position if disputes arise later.

Where there’s no contract, the question of liability will usually be determined by the law of negligence, with established legal principles applied to the analysis and adapted where necessary to the novel factual situations that AI may create. Essentially, this means asking whether the defendant owed a duty of care, whether they fell below the required standard and whether that failure caused foreseeable harm.

Vicarious liability

The UKJT takes the view that AI doesn’t have legal personality and is therefore not a ‘person’ in the eyes of the law, meaning that no one can be vicariously liable for the actions or failures of an AI system. However, an employer can be held vicariously liable where AI-related harm arises because a human employee acted wrongfully while using AI. In other words, the liability still flows through the human, not the machine.

Professional liability

A professional’s obligation to exercise reasonable skill and care in carrying out their duties extends to their use of AI.

Measured against that standard, a professional could be found negligent for using AI inappropriately, choosing an unsuitable model, failing to carry out proper due diligence, or failing to test and validate AI outputs. Perhaps more strikingly, the statement also concludes that a professional could be liable for failing to use AI at all, where a competent member of their profession would have done so.

This is an area that the legal profession is watching particularly closely. As AI capabilities develop, the expectation that professionals will use these tools is only likely to grow and it remains to be seen what reasonable competence looks like in the age of AI.

Liability in the absence of fault

The general position in English law is that, without fault, the risk of harm lies where it falls. However, there are exceptions. Under the Consumer Protection Act 1987, strict liability – meaning liability regardless of fault – can be imposed for death, personal injury or property damage caused by a defective product. As the law currently stands, this would only apply to AI where it’s incorporated into a physical product, since the Act applies to ‘goods’ rather than pure software. The Law Commission has indicated it intends to review product liability law, which may change this position in future.

Causation

The UKJT takes the position that in general the standard ‘but for’ test for causation (ie, would the harm have occurred but for the defendant's actions?) should be applicable in the context of AI. While AI's complexity can make it difficult to understand precisely why a particular outcome occurred, English law has mechanisms for addressing evidential difficulties and these challenges aren’t unique to AI.

The question of whether someone who develops AI can be liable for its misuse by a bad actor or for autonomous acts of the AI is acknowledged to be highly fact specific. However, the UKJT takes the view that someone who develops or deploys AI is unlikely to be held liable for misuse by a bad actor unless the AI in question was obviously dangerous, or the person could have prevented the misuse but failed to do so. That said, a developer or deployer of an AI system would likely be liable for harms caused by the AI acting autonomously, unless acts of the kind in question were unforeseeable.

The UKJT also considers the application of the partial defence of contributory negligence, concluding that in some cases this might be available to reduce damages. In this regard, a non-commercial user of AI might generally be less likely to be found contributorily negligent than a commercial user, but this will depend on the factual context.

What about AI chatbots that make false statements?

AI chatbots are a key focus. The UKJT concludes that a person may be liable for a false statement made by an AI chatbot if the statement was made negligently or fraudulently, or if it’s defamatory. Liability for deceit could also arise, but only where the developer or user intended the AI to produce the statement and for someone to believe it was true.

The UKJT suggests that liability would generally be established if a legal person held out an AI chatbot as communicating on their behalf, or if there’s an express or implied representation that the chatbot’s statements are correct.

Liability can also arise for defamatory statements made by AI where a person is deemed in law to be a publisher of the AI’s output. The extent to which those within the supply chain will be considered publishers will be fact specific, but those who exercise manual review over the output before it’s published will be liable as editors.

The context of statements (eg, clear warnings that words are AI generated and may contain hallucinations) will influence the extent to which they are defamatory.

What does this mean for AI claims?

The UKJT’s statement is not legally binding, nor is it intended to constitute legal advice. The UKJT acknowledges that the nature and breadth of the subject matter are such that a comprehensive analysis of the legal landscape is impossible. A number of areas are also specifically out of scope. These include various bases of legal liability (such as economic torts and competition claims) and areas of law such as the interaction between AI and IP rights, data protection, and issues that arise specifically out of the use of AI by public authorities. Legal issues that aren’t concerned specifically with harms, such as how the law of contract formation applies in the context of AI, are also not addressed.

For this reason, the statement is unlikely to be the final word on AI liability. However, it provides a useful framework and answers business interest in managing uncertainty. For now, it’s likely to be an influential reference tool.

For law firms and their clients, the key takeaway from the statement is that liability for AI harms will be determined largely by existing legal principles, and that contractual arrangements may be the first line of defence. Companies deploying AI should review their contracts, supply chain arrangements and internal governance now, rather than waiting for the law to catch up – a process that will inevitably take time.

The range of the UKJT’s statement is an important reminder that AI liability is not a niche specialism. It cuts across contract, tort, product liability, professional regulation and more. As AI becomes embedded in more industries, lawyers across many practice areas will need to be able to advise clients on these issues. Following this rapidly developing legal landscape closely is a valuable head start.

Charlie Morgan is an international arbitration partner at HSF Kramer .