Federal Court of Australia Generative AI Practice Note: Points to Note on Court Documents, Evidence and Confidential Information

On April 16, 2026, the Federal Court of Australia published the Use of Generative Artificial Intelligence Practice Note (GPN-AI) in the name of Chief Justice Mortimer. The practice note sets out the Court's expectations regarding the use of generative AI for everyone involved in the Court's proceedings. It identifies situations requiring particular care in preparing submissions and evidence, in using generative AI for affidavits and expert reports, and in handling confidential information, as well as situations in which disclosure of use is required. This article organizes the legal nature and content of the practice note based on the primary source and then considers points that legal staff at Japanese companies can draw on. Note that the practice note is a document concerning litigation practice in the Federal Court of Australia and does not apply directly to Japanese civil procedure.

What Was Published: Nature of the Document and Scope of Application

GPN-AI is a General Practice Note issued by the Court under the Federal Court of Australia Act 1976. It is not a statute or rule itself, but a document in which the Court sets out its expectations for the conduct of proceedings. It applies to everyone who appears before or files documents with the Court, including litigants (including self-represented litigants), witnesses, and third parties ordered to produce documents by subpoena or otherwise.

The Court had previously set out interim expectations (responsible use, and disclosure when requested by a judge or registrar) through the Chief Justice's statement of March 28, 2025 and the Notice to the Profession of April 29 of the same year. GPN-AI is the Court's first comprehensive practice note, developed after internal and external consultation during that period. It is intended to be read together with the Central Practice Note (CPN-1) and the Technology and the Court Practice Note (GPN-TECH), and where additional orders are made in individual proceedings, those orders take precedence.

What the Practice Note Provides: Where Use Is Permitted and Where It Is Restricted

GPN-AI does not prohibit the use of generative AI across the board. While recognizing its potential to improve efficiency, reduce costs and enhance access to justice, it is structured to require responsible use in accordance with existing legal and professional obligations.

For originating documents, submissions and other documents filed with the Court, the practice note expressly states that generative AI may produce fictitious cases and citations, inaccurate legal information and factual errors, and expects that, for documents that must bear the name of a responsible person, the responsible person has confirmed the following: the facts are based on matters the party reasonably believes can be proven; the cases and legal authorities cited exist and support the propositions for which they are cited; the evidence cited exists in the material before the Court and is likely to be admissible; chronologies are accurate; and lists of documents comply with the form prescribed by the Federal Court Rules.

More careful treatment is required for affidavits, witness statements and expert reports. Because affidavits and witness statements carry a representation that they reflect the person's own recollection, knowledge and experience, any use of generative AI must be consistent with that premise. Expert reports must reflect the expert's own opinions and reasoning, based on the expert's overriding duty to assist the Court impartially, and separate disclosure obligations are also imposed under the Expert Evidence Practice Note (GPN-EXPT).

Specific cautions are also given on the handling of confidential information. The practice note states that entering information into publicly available generative AI tools may result in the information leaking without the user knowing where it is stored, how it is used, or who can access it, and that careful consideration is required as to whether the following may be entered into generative AI tools: information subject to the Court's confidentiality or suppression orders, privileged information, information that may not be used for purposes other than the proceedings because it came from another party or a third party, and other information that is confidential without the person's consent. The practice note also points out that even with generative AI tools operated in a closed environment, diverting output to another purpose may amount to a similar breach of obligation.

Disclosure Obligations and Consequences of Non-Compliance

GPN-AI does not impose a uniform disclosure obligation on all documents; it establishes a general framework under which the Court may require disclosure on a case-by-case basis. For evidentiary documents, however, a separate disclosure obligation is provided: disclosure is mandatory where a witness has used generative AI to summarize or analyze information on which the witness bases facts or opinions, where images, video, audio or the like filed with the Court were created using generative AI, and where generative AI has otherwise been used in a manner that could reasonably affect the admissibility of evidence or the Court's use of it. Disclosure is to be made by briefly stating at the beginning of the body of the document where and how generative AI was used, and the stated purpose is to prevent the Court and other parties from being misled about how the evidence was prepared.

As for the consequences of using generative AI in a manner contrary to the practice note, the practice note states that costs orders or other adverse orders may be made, and that issues may arise as breaches of legal and professional obligations. It does not expressly provide for criminal penalties or for evidence to be rendered invalid; I think it is necessary to understand accurately that the matter is structured to be dealt with through case management, costs rulings and professional ethics.

In addition, the Supreme Court of New South Wales issued Practice Note SC Gen 23 in January 2025, which imposes restrictions going further than the Federal Court's GPN-AI, such as prohibiting the use of generative AI to generate the content of expert reports without the court's leave. The fact that rules differ from court to court even within Australia is something to check in matters spanning multiple jurisdictions.

Implications for Japanese Practice

GPN-AI is a practice note covering proceedings in the Federal Court of Australia only and does not directly govern the conduct of attorneys or parties in Japanese civil proceedings. However, the issues of where responsibility lies for checking the accuracy of documents, disclosure of generative AI use in the process of preparing evidence, and control over the entry of confidential information can arise in substantially the same way in litigation in Japan. Where an attorney uses generative AI to prepare drafts of briefs or witness statements, it is understood that, from the standpoint of the duty of good faith under the Basic Rules on the Duties of Practicing Attorneys and the duty of care of a prudent manager under the mandate agreement, the attorney cannot escape the duty to personally perform the final check of whether the cases and statutory provisions cited exist and support the propositions, and whether the facts are based on information and evidence obtained from the client. This overlaps with the issues covered in Do Not Use Generative AI Answers As-Is: Why a Human Must Always Do the Final Check and Always Check Article Numbers and Sources Yourself: A Pitfall of Generative AI.

The items that corporate legal staff should check going forward are as follows.

• Ask the outside law firms handling your litigation whether they use generative AI in preparing briefs, witness statements, expert opinions and the like, and what their policy is on disclosure to the client.
• Where generative AI is used in-house for litigation-related documents (chronologies, evidence tables, internal investigation reports and the like), establish operating rules that define, by document type, the extent to which information subject to confidentiality orders or information disclosed by the opposing party may be entered.
• When involved in disputes in which courts in Australia, New South Wales or other foreign jurisdictions are involved, ask local counsel whether the local court handling the matter has issued a practice note on restrictions on the use of generative AI or disclosure obligations.

The division of roles between generative AI and attorneys itself is organized in What Is an AI Lawyer? How to Divide Work Between Attorneys and AI in Corporate Legal Practice in the Generative AI Era. For the practical division of work when a law firm incorporates generative AI tools into its operations, not limited to litigation, Why We Use Codex for Legal Work may also be helpful.

Developments to Watch

The text of GPN-AI states that the Court plans to hold a symposium within the next few months to discuss the challenges and benefits of generative AI. In addition, according to commentary from law firms and others, the Supreme Court of Victoria issued its own practice note and guidelines for judges in May 2026, but this has not yet been confirmed in primary sources.

• Whether the symposium announced in the text of GPN-AI has been held, and what was discussed there
• The content of documents referred to as related materials of GPN-AI, such as the "AI Transparency Statement"
• Whether the Supreme Court of Victoria and other Australian state courts have issued similar guidance, and its consistency with the Federal Court's approach
• Cases in which GPN-AI has been applied in the Federal Court since its publication, and rulings such as costs orders

These are matters that could not be confirmed as of the date of writing or that require ongoing monitoring, and once confirmed information becomes available, checking it against the content organized in this article will be the starting point for a response.