The Pulse
England and Wales Judiciary Sets Guardrails for Judicial AI
The England and Wales judiciary published a position statement on October 9, 2026, setting conditions for AI tools used in judicial work. It requires human review of AI outputs and limits use to approved tools with specified data protection
AI.info Team · Written with AI ·
England and Wales’ judiciary is testing AI to help with tasks such as summarising documents and building chronologies, while insisting that no system may decide a case or replace a judge’s reasoning. A position statement published October 9 sets conditions for judicial AI use, including mandatory human review and approval of every output used in judicial work.
The statement describes a cautious approach to adoption, not a new rule for litigants or a permission for AI to make judicial decisions. It guides senior judges working with the Judicial Office, HM Courts and Tribunals Service and the Ministry of Justice on training, pilots, governance and transparency.
Judges retain responsibility for every decision
The statement draws a firm boundary between AI assistance and judicial authority. “Judicial decisions will be taken by a judicial office holder. AI may assist, but it must not displace reasoning, evaluation of evidence or the exercise of judicial discretion,” it says.
It also requires the judicial office holder responsible for work to review, verify and approve every AI output used in that work. That person remains accountable, the statement says, while judicial independence, impartiality, fairness, confidentiality and security must be preserved.
The distinction matters because the judiciary is not ruling out AI support. Instead, the statement ties any such use to human oversight and says AI tools must not determine a case’s outcome or take the place of judicial analysis. It frames public confidence and open justice as part of the safeguards, alongside accuracy and security.
Pilots test seven kinds of assistance
The judiciary says its pilots have focused on lower-risk tasks and tested secure tools in controlled settings before any wider deployment. The work has examined summarisation, transcription, redaction, writing in accessible language, administrative and leadership tasks, chronology-building and document analysis.
Those pilots have also tested tool reliability, how much human checking is needed, how sensitive material is handled and which safeguards would be required for operational use. The statement does not give performance results or identify the tools used in the pilots, so it does not establish that any particular task has proved reliable enough for routine use.
The new document follows judicial AI guidance first issued in December 2023 and updated in April and October 2025. The judiciary says those earlier updates addressed risks including hallucinations and bias, and reinforced that judges remain responsible for material produced in their name.
Only approved tools can be used
For judicial purposes, the statement permits only approved AI tools. It does not publish a list, citing the pace of technological change, but says approved tools must include UK data residency, enterprise data protection and disabled web access.
New tools require approval, evaluation and ongoing review before wider deployment is considered. A central steering group chaired by the Lead AI Judge is designated as the sole route for oversight of AI development, pilots and deployment; proposed new uses must also be brought to judicial leadership.
Training and review sit alongside safeguards
The Judicial College provides AI training, practical support and guidance through in-person and online sessions, with materials including recordings. The statement says some training covers general skills while other material addresses particular pilots, tools or products.
The steering group will keep the policy under review and may change approved uses, safeguards or governance arrangements in response to evaluations, incidents or emerging risks. For now, the published conditions are clear: AI may assist judicial work, but a judge must check its output, retain responsibility and make the decision.