Federal Circuit and Family Court of Australia issues Practice Direction on the use of AI: what court users need to know

In Australia, early judicial engagement with AI has largely emerged from the FCFCOA, particularly in family law proceedings including involving self-represented litigants. Both Helmold & Mariya (No 2) [2025] FedCFamC1A 163 and Mertz v Mertz (No 3) [2025] FedCFamC1A 222 have become central reference points in discussions about the use of AI, including in relation to legal professional privilege.

In both cases, the FCFCOA made clear that the use of generative AI in preparing court documents can raise serious issues. Importantly, the Full Court observed that:

• inputting court materials into open AI systems may breach confidentiality obligations; and
• may also waive legal professional privilege, particularly where information is disseminated or stored by third-party systems.

These decisions signalled a clear judicial concern about AI undermining fundamental protections over sensitive information. The Practice Direction can be seen as the court’s response to these emerging risks, moving from judicial warning to formal regulation.

Prior to the Practice Direction, practitioners were required to navigate AI risks through existing professional ethical obligations, with practice notes issued by other courts providing non-specific guidance. 

There was no court-specific guidance on how these duties applied to AI. The cases above filled that gap to some extent, but they did so reactively, in response to problematic conduct such as hallucinated authorities and misuse of confidential material. 

The FCFCOA Practice Direction now provides a clear and comprehensive framework grounded in existing legal duties but tailored to AI use. 

It emphasises that AI use must align with fundamental obligations, including integrity, accuracy, accountability, compliance with professional obligations, confidentiality and data security. 

Importantly, the direction stresses that ‘[t]he Courts will not accept a lack of understanding as justification for inadvertence in the use of AI’. Users must understand the risks and limitations of the tools they use, before using them.

The most notable aspect of the Practice Direction is its strong emphasis on protecting confidential, sensitive and privileged information. 

It recognises that generative AI tools may retain and reuse input data, store queries externally and make information accessible to third parties. As a result, the Practice Direction prohibits inputting confidential or sensitive information into public AI tools and mandates an express assurance that any AI system used has appropriate safeguards before uploading such material. 

A strict prohibition applies to discovery or disclosure orders, suppression or non-publication orders and subpoenas. These must not be entered into any generative AI tool unless stringent safeguards are satisfied, including that the information will remain within a controlled environment and not be used to train models.

Court users should also be able to explain, if required, whether and how AI has been used, including what tool was used, how outputs were checked and supervised, and point to alignment with the principles in the Practice Direction. 

The court also takes a cautious approach to evidentiary material. Parties must ensure that:

• affidavits reflect the witness’s own words and knowledge; and
• AI is not used in a way that distorts or fabricates evidence.

This aligns closely with broader concerns about authenticity and reliability in AI-generated content.

The FCFCOA Practice Direction does not operate in isolation. It forms part of a broader, rapidly developing regulatory landscape.

The Federal Court’s Generative AI Practice Note (April 2026) (see our previous article, ‘Federal Court releases Use of Generative AI Practice Note: key guidance for using AI in proceedings’ for more on this) similarly emphasises verification, accuracy and the prohibition on misleading the Court.

Notably the FCFCOA Practice Direction also reinforces the distinction between public and closed AI tools. This largely mimics the recent Supreme Court of Victoria’s Practice Note (May 2026) (see our previous article, 'Supreme Court of Victoria issues AI guidelines for court users and judicial officers' for more on this).

The Victorian Law Reform Commission’s ‘Artificial Intelligence in Victoria’s Courts and Tribunals Report’ (tabled in February 2026) underpins many of these developments, identifying privacy, data security and the risk of privilege waiver as central concerns, and recommending a principles-based approach to regulation.

Viewed together, these instruments demonstrate a consistent and deliberate direction by Australian courts to not prohibit AI but still impose increasingly structured safeguards around its use.