AI in family law: document preparation with care for sensitive data
By the ChamberOne team 5 minute read
Family law runs on paper: bank statements, pension letters, property valuations, schedules of assets, chronologies of a marriage. It is exactly the sort of material that AI tools can sort and summarise. It is also some of the most sensitive information a firm holds, about money, health, children and sometimes safety. The question is not whether to use AI, but which parts of the work suit it and which should stay behind a closed door.
Key points
- AI suits disclosure summaries, draft schedules, chronologies and comparisons, always checked by the lawyer.
- Check sums with a calculator or spreadsheet. Language models can get arithmetic wrong.
- Never enter details that could reveal where an at-risk person or a child can be found, and treat material about children as most restricted.
- Write tool restrictions down and use restricted matters inside the firm for sensitive cases.
Where the time goes in a financial case
In a case involving money, a large share of the work is gathering and reading disclosure. Clients bring bags of statements. Someone has to total the income, list the accounts, find the large transfers and build a picture of the finances. Much of that is routine reading and arithmetic, which is where tools can save hours.
A tool that can read a set of statements and produce a first-pass table of accounts, balances and notable transactions gives the lawyer a starting point. The lawyer's job is then to check it and to ask the questions the table suggests.
Tasks that suit AI
- Summarising long disclosure bundles into a first-pass index of what each document is.
- Drafting a schedule of assets and liabilities from documents, for the lawyer to correct.
- Building a chronology of a relationship or of financial events from the client's notes.
- Comparing two versions of a draft order or agreement and listing the differences.
- Turning technical language in a letter into plain words for a client.
- Preparing a checklist of documents still outstanding, based on what has arrived.
Each of these is a draft for the lawyer to verify against the documents. Arithmetic deserves special attention. Language models can get sums wrong, so total anything that matters with a spreadsheet or calculator and compare.
Information that needs extra caution
Some details in a family file carry risks far beyond ordinary confidentiality. A client's current address, a child's school, or the location of a refuge can put a person in danger if it leaks. Medical records, immigration status and allegations of abuse are highly sensitive. Information about children is a category of its own.
Before any of this goes into an AI tool, decide as a firm what is permitted. A sensible default is that safety-critical details, such as addresses of at-risk parties and the location of children, are never entered, and that material about children is used only in tools the firm has approved for that purpose, if at all.
Set tool restrictions in writing
General instructions to take care are not enough. Write a short rule that staff can apply without thinking.
- Only firm-approved tools, with a contract that prevents the provider using client data to train models.
- No free consumer tools for any family matter.
- Remove names, addresses and identifiers before using a tool, where the task allows, and restore them afterwards.
- Do not enter details that could reveal where an at-risk person or child is living or going to school.
- Treat material about children as the most restricted category, and ask a partner before using any tool on it.
- Check settings so that the tool does not keep inputs longer than necessary.
- Use only firm-approved tools with a contract that stops training on client data
- Remove names, addresses and identifiers where the task allows
- Ask a partner before using any tool on material about children
- Check settings so inputs are not kept longer than necessary
- Use free consumer tools for any family matter
- Enter details that could reveal where an at-risk person or child lives or goes to school
Protecting restricted matters inside the firm
Sensitive family matters are often better restricted inside the firm, so that only the team can see them. This matters particularly where a party is well known locally, or where staff may know the family. Check that any search or AI feature in your system respects those restrictions, so that an assistant in another department cannot surface the file by asking a question.
Drafting with care
AI can help draft the first version of a statement, a letter or an order, but family documents are about particular people in particular situations, and generic text can be actively harmful. A statement that sounds smooth but does not match the client's voice may be challenged in cross-examination. A proposed arrangement about children drafted from a template may overlook the facts that matter most.
Use a draft as scaffolding. Rewrite the parts that touch the client's own account. Read the document with the client, and make sure every sentence is theirs and true.
Working with clients who are in distress
Clients in family disputes are often under strain. Some will use AI tools themselves, and may arrive with advice from a chatbot about what they are entitled to, which may be based on another country's law. Listen without dismissing. Explain what is general and what is specific to their case. Encourage them not to put the other party's private details, or their children's, into public tools.
Tone matters. A summary of a client's finances or history is not only data. It is their life, and it should be handled with the sensitivity you would bring to a conversation across the desk.
Questions to ask before using a tool on a family file
- Where does the text go, who can read it and is it kept?
- Will the provider use our data to improve its models?
- Can we switch the feature off for restricted matters?
- Is there a record of who used it on which matter?
A note on language and plain speaking
Family clients often struggle with the vocabulary of the process. A tool can turn a technical letter into plain words, which is a real help, but the lawyer should read the result against the original and correct anything that changes the meaning. Soft language that hides a deadline or a risk does the client no favours. Say clearly what has been asked of them, by when and what happens if they do nothing.
Where it does not help
AI will not assess credibility, weigh a child's welfare, judge whether a settlement is fair or notice that a client is frightened. Those judgments are the core of the work and stay with the lawyer. The tools can reduce the paperwork around them. Use them in that spirit, and with firm rules about what goes in.
- First-pass index of a disclosure bundle
- Draft schedule of assets and liabilities
- Chronology from the client's notes
- Comparison of two draft versions
- Assessing credibility
- Weighing a child's welfare
- Judging whether a settlement is fair
- Noticing that a client is frightened
This article is general information about the law, legal practice and technology, current when it was published. It is not legal advice. Laws and guidance change: check the sources linked above and take advice on your own circumstances from a lawyer qualified where you are.
Read next
AI contract review: a first-pass checklist for lawyers
What to ask an AI to flag on a first read of a contract, what it tends to miss, and where a lawyer must read every word.
5 minute readWhat AI document analysis extracts well, and what it misses
What AI document analysis does well on parties, dates, clauses and citations, and what it misses, with ways to spot the gaps.
5 minute readAI-assisted engagement letters: what to automate and what stays human
Merge fields and templates handle the routine parts of an engagement letter, while scope, fees and limits of liability need a lawyer's decision.
5 minute read