AI in Family Law: Court Rules, Cases and Tools

What the Court Says About AI

The Federal Circuit and Family Court of Australia’s AI practice direction, PD-AI, applies across both divisions. Paragraph 4.5 says court users ‘must verify the output produced’. Paying for a private tool does not establish accuracy.

Affidavit evidence must express the witness’s own knowledge, experience or opinion in their own words. Under paragraph 4.11, court users other than expert witnesses need not routinely declare AI use unless asked. They must be able to explain its use and supervision. Experts have separate requirements.

Paragraph 5.5 restricts entering disclosure, subpoenaed material and other protected information into generative AI. Its conditions operate together. The information must remain in a closed environment with enforceable confidentiality restrictions, be excluded from model training and be used only for the proceeding.

A paid account or a training opt-out alone does not satisfy those conditions. For the broader differences between jurisdictions, use the AI practice note guide.

Cases Where AI Went Wrong

Reported family law decisions show how fabricated authorities create extra work and professional consequences. They also distinguish what a court established from explanations offered by the people involved.

Mertz & Mertz (No 3)

In Mertz & Mertz (No 3) [2025] FedCFamC1A 222, the Full Court considered incorrect authorities in an appeal’s argument and authority list. Amended documents also raised unresolved questions.

The solicitor said a paralegal used AI to alphabetise cases and group them under appeal grounds. The Court did not accept that account as a complete explanation. At paragraph 12, it said the extent and manner of use ‘remains opaque’. It found that AI use had at least produced erroneous citations.

Paragraph 17 places verification responsibility on the author and those accepting responsibility for the document. The same responsibility covers submissions, footnotes and chronologies. Counsel’s explanation that they had not personally used AI did not remove their obligations.

The Court referred the solicitor and two barristers to professional regulators. A referral sends conduct for consideration. It is not a final disciplinary finding. The solicitor also consented to paying $10,000 for costs wasted correcting AI errors. That payment was separate from the appeal costs following discontinuance.

Mertz therefore concerns responsibility across the people preparing and settling a document. It does not establish that every error came from a particular chatbot or a particular prompt.

Dayal

In Dayal [2024] FedCFamC2F 1166, a solicitor tendered a list and summaries of authorities that did not exist. He acknowledged using an AI research module in legal practice software without verifying its output.

Judge Humphreys accepted his apology and acknowledged his remedial steps. The Court nevertheless referred his conduct to the Victorian Legal Services Board and Commissioner. The reasons explain that the referral was not intended as punishment. The regulator would decide whether further investigation or action was needed.

The failure occurred inside professional software. A legal product label did not make unchecked citations reliable.

Safe Uses for Family Lawyers

A safer use of family law AI starts with a bounded task on material your firm is authorised to process. Summarising the firm’s own documents can save drafting time when the approved tool preserves confidentiality and the lawyer checks the result.

For example, a lawyer can ask for a working chronology from approved correspondence. Keep the source document and page beside each event. If correspondence records an allegation, preserve it as an allegation. A fluent summary must not turn it into an established fact.

Legal Aid NSW’s family law AI guidance stresses factual checking and verification against official legal sources. A party remains responsible for signed documents. Evidence must describe what the witness knows, rather than an account invented to fill missing details.

For authorities, open the judgment and read the passage supporting the proposition. Matching a case name to a database entry is insufficient if the cited passage says something different. Treat another chatbot’s agreement as generated text, not independent verification.

A firm can also use approved AI to simplify a draft client explanation. The lawyer must preserve its legal meaning and advice. Practice administration is a separate choice covered by family law software.

AI and Financial Disclosure

AI can organise financial records, but a narrative answer alone does not show that every transaction was captured correctly. Financial disclosure needs the underlying statements and a clear connection between amounts and their sources.

In an illustrative example, a $4,000 debit from one account and a $4,000 credit to another can be the same transfer. Counting both as spending or treating the credit as income changes the picture. The dates and counterparty details help establish the connection. A matching amount alone does not prove it.

A general document summary can also lose the distinction between a missing statement period and a period with no activity. Those are different facts. The first describes the records supplied. The second describes transactions within records that cover the period.

FinTrace takes over statement conversion and analysis for family lawyers handling financial records. The service consolidates supplied transactions into a searchable, categorised Excel ledger and follows transfers across related accounts. Its written findings report identifies suspicious transactions and gaps in supplied statements.

Every finding cites the source PDF filename and page. That gives the lawyer the transaction evidence behind the explanation without repeating the statement review. The result is useful when a property matter spans many accounts and years, because the ledger brings those records together.

Keep the original statements with the working analysis. Decide what the records establish before turning a financial finding into an allegation in an affidavit or submission.