UK family judge uses AI to summarise judgment for parents

TL;DR:

• Her Honour Judge Hesford has used a secured judicial AI application — based on Microsoft CoPilot — to produce two additional judgment summaries in plain language and bullet-point format for parents with learning and cognitive difficulties.
• The case, Re B – Fact Finding (Use of AI for Judgment Summary for Parents with Learning Difficulties), involved care proceedings over a skull fracture suffered by an eight-month-old child while in her parents’ care.
• The judge confirmed all advocates agreed the summaries were “immensely useful” — the appended sections have been published explicitly to demonstrate appropriate judicial AI use.

The case is one of the first recorded UK examples of a court ruling being explicitly produced in part with generative AI — and one of the first to do so on accessibility grounds rather than for efficiency. The use of AI was not to write the judgment itself, but to make the existing judgment accessible to parents whose cognitive abilities meant they could not engage with the full text.

The judge produced the additional summaries from the full judgment “wholly using the secured judicial AI application, based on Microsoft CoPilot, following prompts to simplify and summarise what had been said.” Both partners in the care proceedings had learning and cognitive difficulties, and were given a simplified-prose summary and a bullet-point version. The agreement of all parties and their advocates was obtained before the AI-derived summaries were appended.

The case behind the precedent

The underlying matter was a five-day fact-finding hearing to establish the circumstances of how a baby (now two years old) suffered a skull fracture at eight months while in her parents’ care. The local authority — not identified in the judgment — has applied for a care order which the married parents oppose.

The judge found that any inconsistencies in either parent’s evidence were due to their confusion, limited awareness and individual difficulties, rather than deliberate deception. She did not find on the balance of probabilities that the injury was deliberately inflicted, but concluded that the care given fell below a reasonable parental standard, leaving both parents within the pool of possible perpetrators. The matter now proceeds to the welfare stage.

Looking forward

The case sits inside a clearly delimited use of AI: post-hoc summarisation of a judge-written judgment, using a secured judicial AI environment (not public ChatGPT), with consent of all parties. That delimitation matters. UK judges have been cautious about generative AI since the Court of Appeal’s 2023 Bandla warning and the Judicial Office’s subsequent guidance, with most permitted uses confined to administrative drafting and document handling. The use of AI for accessibility — making judgments comprehensible to litigants with learning disabilities — is a legitimate procedural innovation, and the explicit decision to publish the AI-derived summaries signals judicial appetite to set transparent norms. The contrast with the Today’s Family Lawyer survey results (covered separately today) — showing 59% of UK legal professionals using unauthorised public AI — is instructive: the bench is modelling responsible AI use just as the practising profession is being caught out for unauthorised use.