Cambridge-Nuffield review finds AI tools deployed across UK justice system are not measuring whether they actually help people seek justice
A research review published in July 2026 by Dr Holli Sargeant of Cambridge University, commissioned by the Nuffield Foundation as part of its Public Right to Justice programme, found that evaluations of AI tools deployed across the UK justice system are measuring operational efficiency but not outcomes that matter to the people the system is meant to serve. The review examined evidence on AI use in administrative, civil, and family justice settings following the rollout of the government's AI Action Plan for Justice in July 2025. Of the 45 AI tools Dr Sargeant examined, only seven had any publicly available evaluation data. Performance claims across the remainder, covering reduced processing times, user satisfaction scores, and improved access, were described as "self-reported and unverified claims, presented without published methodologies". The report identified fairness, comprehension, lived experience, and procedural legitimacy as "the conditions on which the public right to justice depends" and found these were not being measured. A specific finding on litigants in person (individuals representing themselves in court without a lawyer) was striking: analysing over 3,800 queries submitted to a GPT4-powered legal tool by more than 1,000 people, the research found that 70% of queries contained no factual context, reflecting a substantial gap between how ordinary users communicate with AI and how a legal expert would frame the same question. AI to assist courts and tribunals has been deployed most widely, with 12 tools in use for transcription, listing, scheduling, form processing, and case-management. Civil law accounts for eight tools, and family law for six.
Why this matters
The review identifies a structural accountability gap: the government has deployed AI tools across the justice system faster than independent evaluation frameworks have been established to test whether those tools actually improve access to justice. The finding that 38 of 45 tools have no publicly available evaluation data is a direct challenge to the government's AI Action Plan for Justice and creates pressure for a mandatory independent evaluation regime. The litigants-in-person finding is commercially significant: it suggests that AI legal tools calibrated for lawyer-level query framing may actively mislead unrepresented individuals, raising professional responsibility and judicial fairness questions.
On the Ground
The review creates immediate demand for legal technology governance work: law firms and legal technology vendors advising on or supplying AI tools to courts will face growing scrutiny over evaluation methodologies and published performance data. AI governance policy drafting and regulatory impact assessment memos are the primary trainee-level tasks on matters arising from this review. Firms with public law and administrative justice practices will be advising on the government's response and any future mandatory evaluation framework. A trainee would also be reviewing technology licence agreements and data processing agreements for AI tools deployed in justice settings, particularly around data minimisation, audit trails, and performance-reporting obligations.
Interview prep
Question you might get
“What legal and regulatory implications flow from deploying AI tools in the justice system without independent evaluation of their impact on access to justice?”
Sign up free to see the full answer
A model answer you can lift into an interview — how to frame this story for a partner.
Sign up freeMy notes
saved