AI Regulation Tracker  /  Judicial guidance and soft law

UK Judiciary Refreshes Its AI Guidance for Judicial Office Holders

On October 31, 2025, the Courts and Tribunals Judiciary published a refreshed version of its Artificial Intelligence guidance for judicial office holders, updating the April 2025 document. According to the Judiciary's 31 October 2025 publication, it includes updated discussion of the risks and safeguards around AI use, and ties any judicial use of AI to the duty to protect the integrity of the administration of justice. It is internal judicial guidance, not a statute, and it does not create new obligations for the public or for practising lawyers.

By Anthony Guerriero  ·  The Leveraged Years Regulatory Research  ·  Published October 31, 2025  ·  Last updated October 31, 2025  ·  Judicial guidance analysis

The judiciary of England and Wales does not regulate AI for the country. It regulates how its own people use AI. That is worth stating plainly, because a headline about the courts and artificial intelligence can read as if a new rule now binds everyone. It does not. What happened on October 31, 2025 is that the Courts and Tribunals Judiciary updated the AI guidance it first gave judges in 2023 and revised in April 2025 with a refreshed version.

What the refreshed guidance actually changed

According to the Judiciary's 31 October 2025 publication, the update carries updated discussion of the risks and safeguards around AI use. The publication points to a shared glossary of common terms, so that judges reading the document share the same vocabulary for the tools they are being warned about; to risks that include bias in training data and AI hallucinations, which it describes as output that generates incorrect or misleading information; and to confidentiality, reminding judicial office holders not to enter private information into public AI tools and signposting where to report any inadvertent disclosure as a data incident.

None of that is exotic. It is the same short list of failure modes that anyone who has used a general-purpose chatbot for real work already knows. What matters is who is being told, and how it is framed.

The framing is a duty, not a feature list

Lord Justice Birss, the Lead Judge for Artificial Intelligence, tied the guidance to a single principle. In his words, "The use of AI by the judiciary must be consistent with its overarching obligation to protect the integrity of the administration of justice and uphold the rule of law." He added that the guidance reinforces the personal responsibility judicial office holders have for all material produced in their name.

That second sentence is the one to keep. The tool does not carry responsibility. The person whose name is on the output does. A judge who lets an AI system draft a passage still owns every word of it, hallucinated citations included. That is not a UK-only idea. It is the same standard a US court applies when it sanctions a lawyer for filing a brief full of invented cases, and it is the standard a client applies to a professional who hands over AI-assisted work.

Who it applies to, and who it does not

The guidance applies to all judicial office holders for whom the Lady Chief Justice and the Senior President of Tribunals are responsible, together with their clerks, judicial assistants, legal advisers and officers, and other support staff. It does not apply to the public, and it is not addressed to the lawyers who appear in front of these judges, though they should read it, because it tells them exactly what the bench is now primed to look for. A skeleton argument with a citation that does not check out lands very differently in front of a judge who has just been handed refreshed guidance about AI hallucinations.

Why a guidance document for judges reaches US practice

This does not bind anyone in the United States, and it does not create liability for a US firm. Its use is as a benchmark. Courts are conservative institutions, and when a court system writes down how it expects its own members to use AI, it produces a clean, defensible statement of the standard of care. My read is that the UK judiciary has now given every professional a citable reference point. If you are writing an AI use policy for a law firm, an accounting practice, or an in-house team, you can point to a national judiciary that has told its judges, in public, to treat AI output as their own responsibility, to assume it can be confidently wrong, and to keep confidential material out of public tools. That is a stronger anchor than an internal memo, and it costs nothing to read.

The context behind it is the run of cases in which lawyers filed submissions containing citations that AI tools had simply invented, and courts had to decide what to do about it. This refreshed guidance is the bench arming itself for exactly that. For any professional relying on AI, the practical instruction is the same one Birss gave the judges: you are responsible for what goes out under your name, so verify it before it does.

What to do now

Read the guidance document itself, which the judiciary publishes as a PDF, and treat it as a model rather than a foreign rule. Lift its structure into your own policy: a shared glossary so people know what they are talking about, an explicit warning that AI output can be fluent and wrong, a confidentiality rule that keeps client and private data out of public tools, and a clear line that the human signing the work owns it. Then verify AI-assisted citations and facts against the primary source every time, because the failure the courts keep seeing is the one nobody checked.

Browse the full AI Regulation News tracker

Informational analysis for working professionals, not legal advice. Confirm how any guidance, statute, or requirement applies to your situation with qualified professionals in the relevant jurisdiction.