Using AI in Your Family Case
What this Guide is (and isn't)
This guide aims to help you consider the advantages and pitfalls of using AI to help you with your family case. It aims to:
- Help you use AI carefully for low‑risk tasks.
- Warn you about big risks to privacy and accuracy.
- Does not give legal advice or tell you what to put about your specific case.
The Three Golden Rules
- Don’t share confidential info with public AI
Never type private details into a public AI (e.g., names, addresses, children’s details, allegations, health, finances, draft statements, or court orders). Treat it like shouting in a crowded room. - Don’t trust AI for legal facts
AI can “make things up” (called hallucinations). It may invent laws, rules, or cases that sound real but are false. If you use false information in court documents, it can harm your case and may lead to costs orders against you. - You are responsible
AI is not a lawyer and doesn’t understand your family. The court needs your evidence and your words. You must check anything AI suggests with trusted sources.
What AI can help with...
Safer, low‑risk uses
- Brainstorm how to structure a simple email (no names, no case facts).
- Draft a plain, generic letter (e.g., a polite request to a school), without private details.
- Ask for plain‑English explanations of general concepts (e.g., “What does ‘hearing’ mean in general?”) — then verify with official sources.
What it is a bad idea to use AI for...
Do not use AI to
- Write legal arguments or tell you what to put in your statement.
- Find, explain or quote laws, rules, or court cases (AI may invent them).
- Draft witness statements, position statements, case summaries, or anything you will send to the court or the other party.
- Summarise private documents by pasting them into a public AI.
Bottom line: Your story and evidence matter most.
Don’t risk your case by trusting a machine.
Privacy: protect yourself
Never paste into public AI:
- Names, addresses, dates of birth, or contact details.
- Your child’s details or school.
- Allegations, medical or social work information.
- Draft statements, orders, reports, or any court document (even if “redacted” — it’s easy to de‑anonymise).
- Bank details, income, or other financial information.
Checking information (verification you must do)
If AI mentions a law, rule, practice direction or case, you must verify it yourself:
- Laws (e.g., Children Act 1989) → https://www.legislation.gov.uk/
- Family Procedure Rules & Practice Directions → https://www.justice.gov.uk/courts/procedure-rules/family
- Court forms and official guidance for the public → https://www.gov.uk/ (e.g. search: “child arrangements order”, “represent yourself”)
- Judicial announcements/practice guidance → https://www.judiciary.uk/
- Cafcass information for families → https://www.cafcass.gov.uk/
- Case Law → https://www.bailii.org
- Information rights & privacy (AI and data protection) → https://ico.org.uk/
Verification you must do
How to verify (simple steps):
- Search the exact title, section number or case report on the official site (above).
- Read the original text, not a summary.
- For a rule, look for the latest Family Procedure Rules and any Practice Directions.
- If you can’t verify it, don’t use it.
Why this matters:
Using false or misleading legal material can seriously harm your case and may lead to a costs order against you.
If you share confidential information (e.g. family court documents) with a public AI, you are likely to be breaking the law and/or be in contempt of court.
“Enterprise” (confidential) AI vs public AI — what’s the difference?
Most people only have access to public/consumer AI tools (ChatGPT, Gemini, Claude, Grok etc). These may keep or use your inputs to improve their models (some offer opt‑outs, but you should not rely on them for case privacy).
Enterprise/business AI (for example, Microsoft Copilot for Microsoft 365) is designed for organisations and typically offers:
- No data used for training the public model.
- Data stays within the organisation’s secure environment and follows existing permissions.
- Encryption/auditing and controls set by the organisation.
Important: Unless you have a business‑grade AI with a clear confidentiality agreement, assume your data is not private in AI tools.
For family cases, treat confidentiality as essential.
Safe vs unsafe prompt examples
Safer (generic, no private data)
- “Please suggest a clear structure for a short, polite email requesting a list of available dates.”
- “Make this sentence more polite: ‘I cannot attend on Tuesday but can do Thursday morning.’”
- “Explain, in general terms, what a ‘hearing bundle’ means (no case specifics).”
Unsafe
- “Here is my draft witness statement about alleged abuse [pastes text] — improve it.”
- “Here is my court order with names and dates — summarise it.”
- “What law supports me spending time with my child? Draft my argument for the judge.”