Australian AI Governance Briefing: Week Ending 19 April 2026

With Parliament still in recess ahead of Budget week beginning 12 May, the judiciary delivered the defining development. Chief Justice Debra Mortimer of the Federal Court of Australia issued the Generative Artificial Intelligence Practice Note (GPN-AI) on 16 April — the first federally binding rulebook for the use of AI in Australian litigation. The note requires disclosure of material AI use in Court documents, reinforces practitioner responsibility for verifying cited authorities, and foreshadows adverse costs orders and professional-conduct referrals for non-compliance. It does not ban AI, but it embeds AI accountability at the level of the filed pleading.

The OAIC added an operational layer to Australia’s overhauled anti-money-laundering framework on 13 April with a new template privacy collection notice for AML/CTF reporting entities, alongside updated guidance that captures the position that reporting entities are no longer required to retain copies of identity documents for record-keeping purposes. The same day, the South Australian Attorney-General’s Department closed its broader deepfakes consultation, moving the state toward a second wave of legislation that extends beyond the nation-leading non-consensual imagery laws that commenced in November 2025.

For compliance teams, the theme of a parliamentary-quiet week is that AI governance has shifted decisively from theory to binding practice. The Federal Court has made AI accountability a pleading-level question, and the OAIC has begun drafting the templates the rest of the economy will be asked to follow as the Privacy Act’s December 2026 automated decision-making transparency rules approach.

The week in review

The Federal Court sets a federal standard for generative AI in litigation

Chief Justice Debra Mortimer’s issue on 16 April of the Federal Court’s new Generative Artificial Intelligence Practice Note, designated GPN-AI, is the most authoritative court-level treatment of AI use issued in Australia to date. Mortimer CJ’s framing was unambiguous: the presentation of AI-generated false or inaccurate material to the Court is unacceptable, and lawyers, self-represented litigants and expert witnesses are each put on notice that the Court will respond with adverse costs orders and, where appropriate, referrals raising issues of compliance with legal and professional obligations.

Four operative obligations do the structural work. First, any person filing a document must take responsibility for confirming that the legal authorities cited exist and support the propositions for which they are cited. Second, where generative AI has been used to draft, summarise, analyse or otherwise materially shape a filed document — or where AI has been used to generate images, video or audio bearing on admissibility — that use must be disclosed at the start of the document, identifying where and how AI was used. Third, affidavits, witness statements and expert reports must reflect the deponent’s or expert’s own recollection, knowledge and reasoning; AI cannot substitute for the witness. Fourth, the note records that entering confidential, suppressed or private information into public generative AI tools may carry serious consequences, echoing guidance from other Australian courts that material typed into public chatbots should be treated as published.

The note is conspicuously not a prohibition. Mortimer CJ expressly welcomes generative AI’s potential to reduce legal cost and broaden access to justice, and a Federal Court symposium on implementation has been foreshadowed for later this year. That posture aligns the Federal Court with the disclosure-based approach adopted by the NSW Supreme Court in Practice Note SC Gen 23 and the South Australian Supreme Court’s generative AI guidelines that took effect on 1 January, rather than the stricter restrictions on AI in expert evidence reflected in some other state-level protocols.

The significance of the note runs well beyond the litigation context. Its four obligations — use the tool responsibly, understand the tool, do not outsource accountability, disclose material use — map directly onto the enterprise AI governance principles that boards are being asked to approve under the Voluntary AI Safety Standard and the National AI Centre’s AI adoption guidance. For general counsel and compliance teams, the practical implication is that any matter that might reach the Federal Court — commercial disputes, regulatory enforcement, class actions — now carries a procedural AI disclosure obligation that sits above internal governance policy. The Federal Court has effectively made AI accountability a pleading-level question; organisations that have treated AI governance as an internal productivity matter should expect it to surface as a litigation-readiness question the next time they are in court.

The note also lands against a backdrop of rising professional-conduct exposure. The first Australian professional-conduct sanctions for AI hallucinations were imposed on practitioners in 2025, and regulatory bodies in multiple states have further matters on foot. Mortimer CJ’s practice note is in one sense a formalisation of a posture the legal regulators have already begun to take — and a signal that the Federal Court is not prepared to tolerate the pipeline of hallucinated authorities that the Chief Justice of the High Court has previously described as turning judges into filters for AI-generated errors.

Primary sources: Federal Court of Australia — GPN-AI Practice Note

OAIC bolts privacy onto the new AML/CTF regime

The OAIC on 13 April published a template privacy collection notice for AML/CTF reporting entities, together with updated privacy guidance for the overhauled anti-money-laundering regime that commenced for Tranche 1 entities on 31 March and extends to Tranche 2 entities — lawyers, accountants, conveyancers, real estate agents, trust and company service providers and dealers in precious metals and stones — on 1 July. The template is designed to help reporting entities discharge APP 5 collection-notice obligations when gathering personal information for customer due diligence, and reflects the drafting philosophy the regulator has been testing across its recent compliance activity: notices should be specific to the entity’s actual activities, written in plain English, and designed to build trust with the individual rather than to manage legal risk.

The commercially significant change buried in the updated guidance is the retention position. From 31 March 2026, reporting entities are no longer required to retain scanned copies or photocopies of identity documents for AML/CTF record-keeping purposes. Reporting entities should take reasonable steps to destroy or de-identify copies of full identification documents once they are no longer needed, retaining only the underlying data points — name, date of birth, document number, expiry, verification outcome. Copies made before 31 March 2026 may continue to be held for the statutory seven-year period.

For banks, financial services firms, real estate agents and the tens of thousands of other entities across the economy now captured by the regime, the implications are concrete. Retention of identity document images has been a persistent source of breach risk, and the updated guidance effectively reframes how customer onboarding flows should be designed, with privacy-by-design integrated into the record-keeping architecture. The template also signals the OAIC’s posture as Australia approaches the commencement of the Privacy Act’s automated decision-making transparency rules on 10 December 2026: notices are to be written for the individual, not for the regulator. For APP entities anywhere in the economy rewriting collection notices ahead of the ADM transparency commencement, the AML/CTF template is the earliest concrete indication of the standard the regulator expects.

Primary sources: OAIC — AML/CTF privacy guidance | OAIC — Template privacy collection notice

South Australia closes its broader deepfakes consultation

The South Australian Attorney-General’s Department on 13 April closed its consultation on the regulation of deepfakes beyond the parameters of non-consensual sexual material. The consultation was the next stage of SA’s deepfakes regulatory program following the commencement in November 2025 of nation-leading laws that criminalise the use of AI or other digital technology solely to generate invasive, humiliating or degrading images or videos of real individuals, with penalties of up to A$20,000 or four years’ imprisonment. The closed consultation extended the policy frame to the wider risks deepfakes pose — scams, blackmail, political misinformation and fraud — where existing laws are imperfectly matched to generative technology.

A report based on feedback received will now be provided to the Attorney-General. The closure places SA at the leading edge of state-level deepfakes regulation and foreshadows further legislation that builds on the image-based offences framework already in force. For national organisations operating across state lines and for cross-jurisdictional deepfakes policy work, the publication of the SA report will be an early indicator of the direction state legislative responses to generative AI are likely to take during 2026.

Primary sources: SA Attorney-General’s Department — Deepfakes consultation | YourSAy — Deepfakes consultation portal

Stories

Federal Court issues first federally binding Generative AI Practice Note

Chief Justice Debra Mortimer on 16 April issued GPN-AI, the Federal Court of Australia’s new Generative AI Practice Note — the first federally binding rulebook for AI in Australian litigation. The note requires disclosure of material AI use in Court documents, requires the responsible person to confirm cited authorities exist and support the propositions for which they are cited, prohibits AI substitution for witness recollection in affidavits and expert reports, and warns that entering confidential material into public AI tools may carry serious consequences. Non-compliance exposes lawyers and self-represented litigants to adverse costs orders and potential professional-conduct referrals.

Source: fedcourt.gov.au

OAIC publishes template privacy collection notice for AML/CTF reporting entities

The OAIC on 13 April published a template privacy collection notice and updated privacy guidance for reporting entities under Australia’s overhauled AML/CTF regime, which commenced for Tranche 1 entities on 31 March and extends to Tranche 2 entities — including lawyers, accountants and real estate agents — on 1 July. The updated guidance confirms that reporting entities are no longer required to retain copies of identity documents for record-keeping purposes, and should destroy or de-identify copies once no longer needed. The template is an early indicator of how the OAIC expects APP 5 collection notices to be written as the Privacy Act’s December 2026 automated decision-making transparency rules approach.

Source: oaic.gov.au

South Australia closes broader deepfakes consultation

The South Australian Attorney-General’s Department on 13 April closed its consultation on the regulation of deepfakes beyond non-consensual sexual material, covering the wider risks of scams, blackmail, political misinformation and fraud. A report on feedback received will be provided to the Attorney-General, foreshadowing further SA legislation that would build on the nation-leading non-consensual deepfake laws that commenced in November 2025. For national organisations, the SA report’s eventual publication will be an early indicator of how state regulatory responses to generative AI are likely to develop during 2026.

Source: agd.sa.gov.au