The Statement
On July 7, 2026, the UK Jurisdiction Taskforce (UKJT) — an industry-led body operating under LawtechUK — published its Legal Statement on Liability for AI Harms, the product of a public consultation that ran from January 14 to February 13, 2026 (LawtechUK). Drafted by a team led by Matthew Lavy KC, Richard Munden, Lucy McCormick, Iain Munro, Isabel Barter and Jacob Turner, with input from a wider expert group including Professor Ryan Abbott and Professor Sarah Green, the statement's core finding is blunt: English private law, as it stands, is capable of resolving the great majority of AI-caused civil harms without a bespoke statutory regime (LawtechUK).
The statement walks through negligence, vicarious liability, professional liability, product liability, and the law of false statements — negligent misstatement, deceit, defamation — and finds a home for AI harms in each. Professionals can be found negligent both for using AI carelessly (selecting an unsuitable model, skipping validation, missing hallucinations) and, notably, for failing to use AI once it becomes the standard of care in their field (Burges Salmon). AI systems themselves cannot be vicariously liable — they lack legal personality — but employers remain on the hook for employees' negligent AI use, and parties with non-delegable duties cannot outsource that responsibility to a third-party model developer. Businesses cannot claim "technological neutrality" to escape liability when a chatbot generates defamatory or false output (Legal IT Insider). Foundation model developers, by contrast, will typically not be liable for unforeseeable downstream misuse of general-purpose systems — a meaningful limiting principle for the AI industry.
The Case for a Statute
The strongest argument for AI-specific liability legislation is not hypothetical. Complex AI systems are often genuinely opaque even to their operators, and centuries-old tort doctrine was built around human actors whose conduct could be inspected and whose foreseeability could be assessed in relatively linear terms. Claimants harmed by an AI system frequently cannot show what went wrong inside it, who controlled the relevant training or deployment decisions, or whether the harm was foreseeable to a developer who never anticipated the specific downstream use. An EU-style liability directive — with a rebuttable presumption of causation and mandated disclosure of technical documentation from AI companies — would shift that evidentiary burden onto the parties who actually hold the information, a real and defensible access-to-justice concern, especially against well-resourced foundation model developers.
That was, in fact, the EU's own approach until it collapsed. Brussels proposed an AI Liability Directive alongside the AI Act, but by February 2025 the European Commission concluded there was "no foreseeable agreement" among member states and co-legislators, and formally withdrew the proposal, with the notice published in the Official Journal on October 6, 2025 (IAPP). The result is that the EU now has an elaborate ex-ante compliance regime in the AI Act but no harmonized ex-post liability rule — victims of AI-caused harm in the EU must instead navigate 27 divergent national tort regimes, precisely the patchwork the directive was meant to prevent.
Why the Common-Law Bet Is the Better One
Against that backdrop, the UKJT's answer is less a rejection of the EU's underlying concern than a different mechanism for addressing it: let judges apply and adapt existing doctrine case by case, rather than freeze a set of rules into a statute written for AI systems that will look different in five years. Sir Geoffrey Vos, England's Master of the Rolls, framed this explicitly as a competitive claim — that the statement gives "legal certainty and legal predictability" at a moment when "other jurisdictions are still debating how to regulate AI liability," positioning English law as the preferred choice for AI-related contracts (5RB). Matthew Lavy KC put the underlying premise more simply: "English law has long been capable of rising to the challenges of novel technologies."
That premise deserves scrutiny rather than automatic acceptance. Common-law adjudication only works if enough cases actually reach judgment to build a coherent body of precedent — AI disputes, like most civil claims, mostly settle, which risks leaving genuinely novel questions (multi-agent causation chains, autonomous-system foreseeability) under-litigated for years. The statement itself flags one concrete gap worth watching rather than dismissing: product liability law currently only reaches AI embedded in a physical product, leaving pure software largely outside its scope, a gap the Law Commission is reviewing separately.
Even so, the UKJT's approach is the more proportionate one. A statute calibrated to today's AI architectures is liable to misfire against tomorrow's, and the EU's own experience — years of negotiation collapsing into no harmonized rule at all — is a cautionary tale about the cost of trying to legislate liability for a technology still in flux. Common-law flexibility isn't a loophole for AI developers; the statement's own findings on professional negligence and non-delegable duties show claimants already have real routes to redress. For other common-law jurisdictions weighing whether to legislate or wait, England's answer is: don't rush a statute the courts can outgrow.