AI in the family courts: A tool for accessibility, not substitution

The recent decision in Re B – Fact Finding (Use of AI for Judgment Summary for Parents with Learning Difficulties) saw a family judge use artificial intelligence to assist parents in understanding a complex judgment.

In Re B, HHJ Hesford appended two additional versions of her judgment – one simplified narrative and one in bullet-point form – specifically to assist the parents involved in the care proceedings, who both had learning and cognitive difficulties. The summaries were generated using a secure judicial AI system based on Microsoft Copilot, drawing directly from the full judgment and guided by prompts to simplify and explain the court’s reasoning.

This provides a useful illustration of how AI can be deployed responsibly within the family courts. While much commentary on the use of AI in law has focused on its risks, this example demonstrates a more constructive and careful approach: enhancing accessibility without comprising on judicial integrity.

Crucially, the AI did not generate new content or introduce external material. Instead, it processed the judge’s own approved reasoning and translated it into a more accessible format. All the parties and advocates involved agreed that the summaries were “immensely useful”, underlining the practical benefit of this approach when deployed sensibly and in a carefully managed setting.

This measured use of AI stands in stark contrast to the growing number of instances where generative AI has been used improperly in legal practice – courts across England and Wales have already had to grapple with cases involving fabricated authorities, AI hallucinations, and inaccurate submissions generated by publicly available AI tools. In one widely reported example involving a High Court case, dozens of fictitious case citations were put before the court, prompting serious concerns about the administration of justice and public confidence in the legal system if AI continued to be misused by officers of the court.

However, the case of Re B demonstrates a different approach to how AI can be implemented into legal practice in a beneficial way. The AI system was confined to the judgment itself and the output was limited to producing a summary only. There was no opportunity for the AI to “invent” authority or introduce error.

Such an approach can lend itself to improving the experience of many within the family courts. Family proceedings often involve litigants in person who face barriers to understanding complex judgments but HHJ Hesford’s approach reflects a broader principle: that the law must be upheld, but must also be understood.

By leveraging AI in this way, the court addressed a frequently occurring issue in family law – how to ensure parties fully grasp the outcome of proceedings that will profoundly affect their lives. This is not a novel problem, but AI provides a new mechanism for addressing it efficiently and consistently.

The legal profession is increasingly engaging with AI but with this comes additional challenges, and both regulators and the judiciary are emphasising the need for caution. The courts have made clear that legal professional remain responsible for the accuracy of any material they present, regardless of whether AI has been used in its preparation.

This dual trend of rapid adoption alongside heightened scrutiny means that the profession must develop a clear understanding of how to implement AI into its practice appropriately. The key lesson from this case is that AI is most effective when it augments, rather than replaces, legal expertise.

How we can help

At Cripps, we recognise that family proceedings can feel overwhelming, particularly where decisions are complex or emotionally charged. Our focus remains on ensuring that advice is not only legally robust, but also clearly explained and practical. Where appropriate, we continue to explore tools and approaches that support client understanding, while maintaining the careful oversight and professional judgment that such matters require.