AI for family firms: risk or reward?

It is unquestionable that the increasing use of AI by both legal practitioners and clients is currently growing and will continue to impact family law. This can be daunting for those who do not feel adept at navigating the numerous, rapidly-evolving technologies and the implications/risks of using them. Conversely, this emerging technology creates real opportunities for family practices of all sizes.

The Courts and Tribunals Judiciary published very helpful Guidance for Judicial Office Holders on 31 October 2025 which emphasises the reality of AI use in the court system, while emphasising that individuals must remain responsible for everything they place before court.

Key benefits

AI tools are capable of summarising vast amounts of information quickly, and streamlining a variety of administrative tasks such as helping to prepare agendas, taking and summarising meeting notes, drafting correspondence or applications and writing presentations. This can free up practitioners for more valuable tasks such as strategising and negotiating, and can assist with keeping client costs proportionate. All such work should be checked carefully to ensure its accuracy.

AI can also be used for case or legislative summary, if using appropriate legal technologies. This is treated with caution in the October 2025 guidance, which suggests that while legal research or analysis via AI may produce useful “reminders” of material, source material should be carefully checked by the user to ensure the correct authorities are being drawn on.

Key risks

AI hallucinations

In the joined immigration cases of UK v Secretary of State for the Home Department [2026] UKUT 81 (IAC), the Upper Tribunal (Immigration and Asylum Chamber) dealt with the issue of AI hallucinations: these are incorrect or misleading results generated by AI tools. The cases contained issues around drafting of court documents using AI. The tribunal noted that where a regulated legal practitioner had not verified citations generated by AI tools, a referral to the regulator would be appropriate (in this case unnecessary as the individual had self-referred).

In Anthony Malcolm Cork & anor v Mark Smith [2026] EWHC 1199 (Ch), an insolvency case, Pinsent Masons were found to have twice misled the court, firstly by providing references containing AI hallucinations and secondly by using AI to generate an explanatory letter which did not remedy the mistakes. In this case, a supervising solicitor and partner were found to have failed to supervise the junior fee earner sufficiently, despite the partner not being aware AI was being used – the firm has since self-referred to the Solicitors Regulation Authority. It is therefore incumbent upon every practitioner to ensure the veracity of the material they put before court, including that of juniors they supervise.

Client confidentiality

In UK v Secretary of State for the Home Department the tribunal also considered that uploading of confidential emails and documents to ChatGPT (an open-source AI tool) constituted placing that material into the public domain, thus breaching client confidentiality and waiving legal privilege. The judge considered this conduct should be referred to the Information Commissioner’s Office and might warrant an SRA referral. Had a closed AI tool been used, this issue may have been avoided and practitioners should be particularly mindful in an area such as family law where the nature of evidence is often inherently sensitive in nature.

Other court users’ practice

Practitioners should be mindful that even if they have not used AI tools, others may well have. Points to look out for include:

• Fake/incorrect material: it may be necessary to ask questions of other parties, particularly if litigants in person, to ascertain whether they have used AI tools and if so, how they have checked accuracy. Practitioners should also be mindful of the use of “deepfakes” created by AI which users may purport to be legitimate documentary evidence. Some tell-tell signs of AI use may be:
  • Unrecognisable citations or those which appear to be from another jurisdiction.
  • Inconsistent case law
  • Inconsistent submissions with your understanding of the practice area
  • Unusual spellings of words/American spellings
• White text: – this is a practice where text is concealed in white within documents so that the AI tool will review it, but a human will not see it. This can lead to inaccurate summaries or conclusions made by the tool.

[email protected]