AI: THE NEW FOURTH MEMBER OF THE BENCH?

Having been retired from the magistracy for over a decade this blog has by default morphed into more of a commentary on general issues than was the case in years gone by.  Freedom to observe and opine on situations that are highly critical of the Ministry of Justice goose and its underwritten goslings eg His Majesty`s Courts and Tribunal Service, The Criminal Cases Review Commission, police governed by the Home Office [Met Police excepted] or the Judicial Conduct Investigations Office which reports directly to the Lord Chancellor  is something impossible for my active former colleagues without incurring the wrath of god.  This omerta by all active judiciary  is worthy of the originators of the term; various groups, political and criminal, in 19th century Sicily and southern Italy. Amongst the topics one doesn`t learn from media, mass or social, is the effect on the judicial system of AI. This post is an initial attempt to throw just a few lumens on this subject.

In 2025  The Ministry of Justice published its AI Action Plan for Justice involving inter alia courts, in an effort to make justice faster, fairer, more accessible whilst protecting human rights, the rule of law and judicial independence.  The purpose was that AI was to assist but not replace judicial decision making.  The plan, we are told, applies to all judicial office holders, clerks, judicial assistants, legal advisers/officers and other support staff.  It would appear that AI would [will?] be used as it is currently used eg with Trip Advisor summing up hotel reviews which in the court scenario would mean summarising lengthy case papers or identifying key words, phrases or sentences. Transcription of speech to text and translation of texts would be just an extension of what is already widely in operation in all walks of society.  

What is not intended is to remove from the human bench the decisions as to whether a defendant is guilty; whether a witness is telling the truth; what sentence should be imposed; whether a defendant should be remanded in custody; whether an individual is dangerous; what findings of fact a bench should make. The MOJ says its approach must preserve human decision making, legal rights and judicial independence. How long this stance can be definitive only time will tell.

In June 2026 the government announced that it is developing AI legal assistants and other AI applications to tackle court delays.  From my own experiences as a user of county courts the administration of routine functions is, to mix metaphors, in the steam age.  The current position is that AI cannot replace the exercise of human judgment.  

What is significant is what has changed since I retired.  Court attendees can manipulate text, images, videos and recordings.  The President of the King's Bench Division recently warned that AI means fabrication which previously required specialist expertise can now potentially be produced cheaply with a smartphone. He specifically identified forged email chains, synthetic voicemail, manipulated CCTV, fake social media exchanges and false expert looking reports.  Bearing in mind that the bench`s function is to find facts, fake evidence may be accepted as genuine and genuine evidence may be dismissed as fake. 

Solicitors and barristers are already able to use AI for the benefit of their clients in a manner similar to the above listed functions outlined by the MOJ. The Civil Justice Council has been examining AI generated pleadings, advocacy documents, witness statements and expert reports. Its June 2026 findings say there is presently substantial support for allowing professional legal drafting under existing professional responsibilities but witness statements remain an area of particular concern because AI can reshape or embellish a witness's recollection. It can potentially influence how a witness's account is drafted. This is of particular concern to senior judiciary. At its annual strategy meeting in January 2025 the Civil Justice Council (CJC) agreed to set up a working group examining the use of AI by legal representatives for preparing court documents.

For a legal advisor in the arena of a magistrates court the use of AI carries equally grave responsibilities. AI can produce a perfectly convincing looking citation to a case which does not exist or misstate what a genuine case decided.  During my career on the bench when a LA gave an option or opinion we could see for ourselves the words on a printed page and exercise our task of judgement and/or interpretation of those words within the law.  Currently magistrates must not allow an AI generated answer to become an unseen source of legal advice just as in my time references eg to Google were forbidden. 

Referring once again to Trip Advisor where AI sorts and categorises thousands of reviews on destinations, hotels and restaurants etc into simplified easily digestible results  senior judges now recognise its potential to help with case management and the handling of very large volumes of information.  However the downside to such abbreviation is that some important facts will be omitted or be overlooked by those reading the shortened document. 

For magistrates just as outside referring was forbidden so it is with AI. The bench cannot ask Claude for advice. Until now a bench was faced with the traditional question: do we believe this evidence? Now the question is  can we be confident that this evidence is what it purports to be?

When I first wrote about the magistrates' courts in 2010-2013 {at another site and now published  at   https://amagistratesdiaries.blogspot.com/ } , artificial intelligence was scarcely a consideration. When I retired in 2015 it was still science fiction. In 2026 it is becoming impossible to ignore.  But would I now feel out of place, out of my depth in the middle chair of my inner city bench?  The basics are still unchanged.  The bench still has to listen to the evidence, assess the credibility of witnesses, apply the law with the assistance of the legal advisor and decide whether the prosecution has proved its case beyond reasonable doubt. Perhaps in the not too distant future AI might be asked for its opinion  but for the foreseeable future it cannot make that decision of whether guilt has been proved beyond a reasonable doubt.  However it is the credibility of the evidence presented to the court which raises the possibility of its origin. AI can produce remarkably convincing photographs, documents, emails, voices and videos which never existed. It can also alter genuine material. The technology is becoming cheap, accessible and increasingly difficult to detect with the naked eye or ear.  The corollary is that genuine evidence might be dismissed as "AI generated" when it is not.  As mentioned above most worrying perhaps is the potential influence of AI on witness statements. If a witness's account has been drafted, polished or reshaped by AI where does the witness's own recollection end and the machine's contribution begin?

When Sentencing Guidelines were introduced in 2004 there was a split in judicial opinion as to their worthiness and a fear that the discretion of sentencers would be reduced if not eliminated.  The practice of structured decision making at all criminal courts was eliminated. As is now demonstrable by the many sentencing pronouncements made on live TV of major criminal offenders the steps to a sentence are almost robotic in their presentation and in my opinion are leading inevitably to a position where, when consulted, the computer says yes.  Although the limits of AI currently seem to be constructed in steel reinforced concrete I do not believe my successors in 2036 will have that inhibition cast upon them.  AI might not replace the magistrate but it might fundamentally change the nature of the task the magistrate is being asked to perform.