Free, public AI tools like ChatGPT can be useful for general questions, but they are not a safe place to discuss the details of your family law matter. Anything you type may be stored and reviewed, and uploading confidential documents means losing control over them. Sharing your lawyer’s advice with an AI tool could even risk waiving the legal professional privilege that protects it. The risks are now so recognised that on 29 May 2026 the Federal Circuit and Family Court of Australia issued a Practice Direction on the use of AI (PD-AI). It applies to everyone involved in family law proceedings, including people representing themselves, and it prohibits putting certain court-related documents into AI tools at all. If you do use AI, keep it to general questions only, with no names or confidential detail.
This guide is part of our series on Safety & Privacy When You Separate.
The temptation is understandable
When you’re separating, questions come at all hours. A free AI assistant that answers instantly, without judgement or a bill, is quite tempting. Many people now type their situation into ChatGPT or a similar tool, or even upload documents and ask it to “explain this” or “write my response.”
We understand the appeal. So do the courts. The Federal Circuit and Family Court of Australia acknowledges that responsible AI use “can have significant benefits for efficiency and access to justice.” But the Court was concerned enough about the risks that it has now issued formal rules, and some of those risks can affect your family law matter, not just your privacy.
Risk 1: Your privacy
When you type information into a free, public AI tool, you generally give up control over it. Depending on the service and its settings:
- What you type may be stored on the provider’s systems.
- It may be reviewed by people (for example, to improve the service).
- It may be used to train future versions of the AI.
The Practice Direction warns that a generative AI tool “has the capacity to retain information entered into it and store every question asked of it,” and that it could use that information to respond to other users. That means that information you enter “could become publicly known.”
Policies vary between products and change over time, so you can’t assume anything you enter is private. If you upload bank statements, court documents or messages, you’ve handed sensitive information to a third party you don’t control.
Privacy after separation isn’t only about what you type. It’s also worth checking whether your phone, accounts or car are sharing your location.
Risk 2: Your legal privilege
Confidential communications between you and your lawyer are generally protected by legal professional privilege (also called client legal privilege). It’s an important protection. Broadly, it means that nobody (including the other side of your matter and the Court) can force you to reveal the legal advice you’ve received.
But privilege depends on the communication staying confidential. If you share your lawyer’s advice with a third party, you can risk waiving that privilege. Pasting your lawyer’s email into a public AI tool, or uploading a privileged document, may amount to disclosing it to a third party. That may put that protection at risk.
The Courts’ new AI rules make the point directly: court users must comply with “confidentiality orders and laws, personal data protection laws, intellectual property laws and legal privilege” when using generative AI tools.
The safest assumption is to treat anything your lawyer sends you as confidential, and don’t feed it into AI tools.
Risk 3: AI gets things wrong
AI tools are confident, but sometimes wrong. They can:
- Give general information that doesn’t fit your circumstances.
- Be out of date (for example, missing recent changes to the Family Law Act).
- Invent things that sound plausible but aren’t correct, including fake “cases” or rules.
The courts see this often enough that the Practice Direction specifically requires anyone using AI to prepare court material to verify that every case, piece of legislation or quote actually exists and says what the AI claims. They also warn that the Court “will not accept a lack of understanding as justification” for AI mistakes.
Australian family law turns on the specific facts of your situation. General AI answers can give false confidence and lead to costly mistakes.
The family courts now have formal AI rules
On 29 May 2026, the Federal Circuit and Family Court of Australia issued a Practice Direction on the Use of Artificial Intelligence (PD-AI). If your matter is (or may end up) before the court, this applies to you. It covers all “court users”, expressly including self-represented litigants, not just lawyers. In plain English, the key points are:
- You are accountable for anything AI helps you produce. If you file a document, you’re responsible for its accuracy. The court can ask you to explain whether and how you used AI, which tool, and how you checked its output.
- Some documents must not go into AI tools at all. Material obtained through court-ordered disclosure or subpoena, and anything subject to a non-publication or suppression order, must not be entered into a generative AI tool (except in tightly controlled “closed” environments most public tools don’t satisfy). Practically speaking, don’t upload the other side’s financial disclosure or any other material to ChatGPT.
- Publishing family law proceedings is a criminal offence. It’s an offence to communicate to the public an account of family law proceedings that identifies a party, witness or child. Because public AI tools can retain and resurface what you enter, feeding identifying case details into one carries real risk.
- Affidavits must be your own words. The court expects affidavit evidence to reflect the witness’s own knowledge “expressed in the witness’s own words”.
- Safety matters. The court warns against entering information into AI tools where disclosure could put someone’s safety at risk, for example a protected address.
None of this means AI is banned. The court’s rules are about responsible use. But they underline the theme of this article: the details of your family law matter don’t belong in a public AI tool.
So how can you use AI safely?
AI isn’t all risk. Used carefully, it can help you prepare. A sensible approach:
- Keep it general. Ask “what is a consent order?” — not questions containing your name, your ex’s name, or the details of your matter.
- Never upload confidential or privileged documents — and never anything obtained through disclosure or subpoena.
- Don’t paste in your lawyer’s advice.
- Turn off chat history if the tool allows it — a precaution the court’s own guidance suggests.
- Use it to prepare questions, not to get answers you’ll rely on.
- Always check anything important with your lawyer before acting.
Think of AI as a starting point for understanding the landscape. But never as your lawyer, and never as a confidant for the details of your case.
The bottom line
For general background, AI can be a helpful first step. For your actual family law matter — the names, the documents, the advice — it isn’t a safe place, and the courts have now said so formally. Protect your privacy and your privilege by keeping confidential information between you and your lawyer.
If you have questions about your separation, talk to someone who is bound to keep your information confidential and can give you advice tailored to your situation. If you’ve recently separated, our free Separation Checklist sets out the practical first steps to take.
Frequently Asked Questions
AI tools can give general information, but they are not a substitute for legal advice and can be wrong or out of date. The family court’s Practice Direction on AI notes that providing legal advice is the role of a lawyer, subject to strict professional obligations. For advice about your situation, speak to a qualified family lawyer.
Yes. Since 29 May 2026, the Federal Circuit and Family Court of Australia’s Practice Direction (PD-AI) applies to everyone involved in proceedings, including self-represented litigants. You’re accountable for AI-assisted documents, must verify AI output, and must not enter certain court-related or confidential material into AI tools.
No. Uploading confidential documents means losing control over them, and may risk the legal privilege that protects communications with your lawyer. Documents obtained through court disclosure or subpoena must not be put into public AI tools at all under the court’s Practice Direction.
Be very careful. The court expects an affidavit to reflect your own knowledge in your own words, and you remain responsible for its accuracy. You may also be asked to explain how AI was used in preparing a document. Get legal advice before relying on AI for court documents.
It’s a protection that generally keeps confidential communications between you and your lawyer private, including the legal advice you receive. It can be lost or waived if the communication is shared with others.
It can. Sharing confidential or privileged information with an AI tool may compromise your privacy and potentially your legal privilege; misusing AI in proceedings can lead to material being disregarded, costs orders or other consequences; and relying on incorrect AI answers can lead to poor decisions.
Keep questions general, never include identifying or confidential details, never upload documents or your lawyer’s advice, turn off chat history where possible, and always confirm anything important with your lawyer.
