Open Generative AI models, such as ChatGPT, are useful tools for efficiency and productivity. However, these tools can be equally dangerous if used incorrectly.
The recent decision of Helmold & Mariya (No 2) [2025] FedCFamC1A 163 is an example of the legal and ethical risks of using Generative AI in litigation - particularly in the context of family law proceedings.
In this matter, the Full Court of the Federal Circuit and Family Court of Australia (Division 1) dismissed an appeal brought by a self-represented party. While the appeal primarily focused on procedural fairness and allegations of bias against the trial judge, the Full Court also focused on the Appellant’s use of Generative AI to draft his Court documents and submissions - with those documents including hallucinated cases (being cases which don’t exist), and cases which do exist but do not actually support the Appellant’s argument.
The Court determined that all litigants, whether legally represented or acting for themselves, have a duty not to mislead the Court or the other parties. Relying on unverified AI can confuse the Court and other parties, unnecessarily complicate the matter, and result in wasted time - and litigants therefore have an obligation to check the accuracy of what is produced before presenting that case.
In a family law context, uploading Court documents to Generative AI may also breach section 114Q of the Family Law Act 1975 - which prohibits the publication of an account of family law proceedings to a section of the public, given that those public models use information they receive to inform their outputs. It may also waive legal professional privilege and expose the litigant to the risk that their confidential instructions and legal advice may be accessed by the other party.
Family law matters are complex and require specialist advice. A key takeaway is that Generative AI is a great tool for efficiency - however it carries significant risks, and should be used with caution. Trust, but verify!
Case link: Helmold & Mariya (No 2) [2025] FedCFamC1A 163
Liability limited by a scheme approved under Professional Standards Legislation
