FCFCOA Practice Direction PD-AI: what lawyers need to know about AI in court

Share:

Family Lawyers reading about the FCFCOA new AI-PD Guidelines

AI in your practice is now regulated, here's what changes

On 29 May 2026, Chief Justice Alstergren issued Practice Direction PD-AI, the first FCFCOA-wide rules governing AI use in proceedings. It applies immediately to all lawyers, litigants and private experts, and covers any tool that performs a task that would otherwise require human intelligence, regardless of whether it is labelled "AI."

What does FCFCOA Practice Direction PD-AI require?

PD-AI requires all court users to verify AI-generated content before filing it, cite only real and current Australian law, protect confidential client information, and be ready to explain their AI use if the court asks. Lawyers who engage private expert witnesses must give each expert a copy of the practice direction and obtain confirmation that they have read it.

"Court users are responsible for keeping up to date on its requirements." - PD-AI, clause 1.5

The direction does not require proactive disclosure of AI use for most documents, but it does require you to be prepared to account for your AI use on demand. Document your process on every matter.

Applies to all AI tools, even those not labelled "AI"

Drafting assistants, smart search functions, transcription services, and document management systems with AI features all fall within the scope of PD-AI. If it performs a task that would otherwise require human intelligence, your obligations apply. (Clause 1.6)

The 6 core principles (clause 3.3)

Any use of AI in FCFCOA proceedings must comply with all six principles. There is no hierarchy, all six apply.

Integrity AI use must support the court's principles of fairness, truth, proportionality, and equality before the law.
Accountability You remain responsible for every document filed. AI involvement does not reduce that responsibility.
Accuracy All AI output must be independently verified before relying on it. Paid or commercial tools are not assumed to be reliable.
Confidentiality and data security Client and case information must remain protected. Understand how your AI tool stores and uses data before entering anything into it.
Safety Do not use AI where it could compromise the safety of a court user or child, including through unintended disclosure of a protected address or medical records.
Education and understanding Lawyers must undertake professional education on AI commensurate with their use of it. The court will not accept a lack of understanding as justification for inadvertent misuse.

Pre-filing checklist for AI-assisted documents

Before filing any document that involved AI assistance, work through each of these steps.

  1. 1
    Verify all case law, legislation, and citations

    Check every reference against AustLII or the Federal Register of Legislation or State or Territory equivalent. AI tools can fabricate case names and section numbers that look credible but do not exist.

    Ref: clause 4.7(c)

  2. 2
    Proofread and adapt output. Do not file verbatim

    Fact-check, edit, and tailor any AI-generated content to the specific matter. Overseas law, generic advice, or outdated authorities must be removed or corrected.

    Ref: clauses 4.7(a)–(b)

  3. 3
    Confirm affidavit content reflects the witness's own words

    A party must ensure that any affidavit is based on the witness's own knowledge, experience, or opinion, expressed in their own words. AI drafting of affidavit content requires particular caution.

    Ref: clause 4.9

  4. 4
    Give expert witnesses a copy of PD-AI and obtain confirmation they have read it

    If you are engaging a private expert witness, this is your obligation, not the expert's. Provide PD-AI and ensure the expert certifies in writing that they have read it before they prepare their report.

    Ref: clause 4.10

  5. 5
    Review your AI tool's terms of service for data use

    Check whether your tool retains inputs, trains on them, or shares them with third parties. Data use practices differ between tools and between subscription tiers of the same tool.

    Ref: clauses 4.18–4.19

  6. 6
    Disable chat history in GenAI chatbots where possible

    If your tool offers the option to disable chat history and web browsing, turn these off to reduce the risk of client data being retained or used for purposes outside your matter.

    Ref: clause 5.4(b)

What is prohibited and what is permitted

Hard prohibitions

  • Do not use AI to alter, enhance, or fabricate images, video, audio, or other media used as evidence. (Clause 4.6(b))
  • Do not enter discovery documents, subpoena material, or information subject to a non-publication order into any GenAI tool unless the tool is a closed environment with enforceable confidentiality terms and the data will not be used to train any AI model. (Clause 5.5)
  • Do not use recording or transcription tools, including AI tools, in court. (Clause 5.6)
  • Do not use AI to perpetrate family violence or litigation abuse. (Clause 4.20)

Permitted and encouraged

  • AI-assisted drafting of submissions, chronologies, and research with human verification.
  • Closed or enterprise AI tools with contractual data protections and no training on inputs.
  • Offering AI-assisted outputs to the opposing party where doing so reduces their manual burden and promotes access to justice. (Clause 4.14)
No mandatory disclosure, but be ready to explain

There is no general obligation for lawyers to proactively disclose AI use in filed documents. However, clause 4.11 requires you to be able to answer, if asked, whether AI was used, which tool, how the output was checked, and how the principles in PD-AI were observed. Keep a brief AI use log for each matter.

Public AI tools vs closed AI tools

Public AI tools

  • Examples: ChatGPT (free tier), Google Gemini, Claude.ai (free).
  • Treat as an open environment. Inputs may be stored, used for training, or accessible to others.
  • Never enter client names, case details, protected addresses, medical records, or subpoena material.

Closed AI tools

  • Examples: enterprise subscriptions with contractual data controls and no training on inputs.
  • Still require verification of all output. A closed environment does not guarantee accuracy.
  • Read the terms of service carefully. Data practices vary significantly between tiers.

Consequences for AI misuse

The court has broad discretionary powers to respond to inappropriate AI use (clause 4.21):
  • Material disregarded: the court may set aside the affected document or evidence entirely.
  • Adverse costs orders: cost orders or other penalties may be made against you or your client.
  • Referral to a regulatory body or Legal Services Commissioner : the matter may be referred to the Law Society or Bar Association.

Lawyers who suspect improper AI use by the opposing party, where it could undermine the integrity of proceedings, should raise it with the presiding judicial officer.

Frequently asked questions

Does FCFCOA PD-AI apply to legal research tools like Lexis+ AI or Thomson Reuters?

Yes. Clause 1.6 states that obligations apply "regardless of whether a product is expressly described as AI, GenAI, smart, assistive or similar." AI-powered legal research tools, smart drafting assistants, transcription services, and AI-embedded document management systems all fall within scope. The label does not matter, the function does.

Do I have to tell the FCFCOA that I used AI to prepare a document?

No general proactive disclosure obligation exists for lawyers (clause 4.11). However, if the court asks whether AI was used, and how, you must answer. You should be ready to explain which tool you used, how you reviewed the output, and how you complied with PD-AI. Keeping a brief AI use log for each matter is the most straightforward way to protect yourself.

What must I do when engaging a private expert witness under PD-AI?

Two steps, both your obligation: (1) provide the expert with a copy of Practice Direction PD-AI, and (2) ensure the expert certifies in writing that they have read it. The expert must then comply with PD-AI when preparing their report or opinion. Failing to take these steps is a risk to you as the engaging lawyer, not only to the expert.

Can I use ChatGPT or other public AI tools for FCFCOA family law matters?

PD-AI does not prohibit public AI tools outright. However, you must not enter sensitive or confidential client information into any publicly available AI system (clause 5.4(c)). In family law matters this is especially significant: disclosing information that could identify a party, witness, or child in proceedings is an indictable offence under Part XIVB of the Family Law Act 1975 (Cth). If you do use a public tool, keep inputs free of any identifiable or sensitive information and verify all output before use.

What happens if AI-generated content I filed turns out to be inaccurate?

The responsibility is yours. PD-AI places the verification obligation squarely on the court user, not the AI tool. If inaccurate AI-generated content is filed, the court may disregard the affected material, make an adverse costs order against you or your client, or refer the matter to the relevant regulatory body such as the Law Society. A paid or commercial tool is not a defence.

Does PD-AI require me to complete CPD on AI?

Clause 3.3(f) requires lawyers to undertake professional education on AI "commensurate with the extent they intend to incorporate it into their practice." It does not mandate a specific number of CPD Points, but it makes clear that using AI without adequate training is a breach of your obligations under PD-AI. Recognised CPD courses on AI in legal practice are the most direct way to demonstrate compliance with this principle.

Practical tip: build an AI use log for every matter

Record the tool used, the task it assisted with, how you verified the output, and (where relevant) that the expert received and certified PD-AI. If the court asks about your AI use, this log is your evidence of compliance. LawCPD's free AI Use Log template is available to download.

Source: FCFCOA Practice Direction PD-AI, issued 29 May 2026 by The Honourable Justice William Alstergren, Chief Justice. This article is an educational summary and does not constitute legal advice.

Stay compliant as AI rules evolve. LawCPD's recognised CPD courses cover AI in legal practice, court obligations, confidentiality, and professional responsibility.

Browse AI and technology CPD courses View all CPD courses

Share: