Artificial Intelligence in the Legal Profession: What the Federal Court’s GPN-AI Means for the Future of Legal Practice
…Unsplash
Artificial Intelligence (‘AI’) is no longer a distant possibility for the legal profession. It is already changing how lawyers research, draft, analyse and manage information. Generative artificial intelligence (‘GenAI’) in particular has moved rapidly from an emerging technology to a practical tool used across legal workplaces. Systems such as ChatGPT, Claude, Harvey, Google Gemini and Microsoft Copilot can produce written material within seconds and perform tasks that once required hours of legal work.[i]
For a profession built around information, the benefits are clear, yet they raise a more difficult question. At what point does using AI to assist legal work begin to affect the responsibilities attached to that work?
The Federal Court of Australia has begun to answer that question. On 16 April 2026, it published its Use of Generative Artificial Intelligence Practice Note, commonly referred to as the GPN-AI.[ii] The Practice Note was published following extensive consideration of the implications of GenAI throughout 2024 and 2025, including consultation with legal professionals, litigants and other courts. It represents the Federal Court’s ‘first comprehensive statement on the use of generative artificial intelligence’ in its proceedings.[iii]
The GPN-AI is notable for what it does not do. It does not prohibit the use of GenAI nor does it treat the technology as inherently incompatible with legal practice. Instead, the Court acknowledges that GenAI has the potential to ‘increas(e) efficiency, reduc(e) legal costs and enhanc(e) access to justice’, while making it clear that these benefits cannot come at the expense of the administration of justice.[iv]
This approach reflects a broader and still developing Australian regulatory landscape. The Victorian Law Reform Commission’s Artificial Intelligence in Victoria’s Courts and Tribunals report similarly rejected both uncritical adoption and blanket prohibition of GenAI. Instead, it advocates for a principles-based approach centred on human oversight, accountability, transparency, protection of evidence and differentiated treatment of higher risk uses.[v]
The emerging position is therefore not about whether AI should be permitted, but whether its use is appropriate to the task, the information involved and the consequences of error. The distinction is not between AI and no AI, but between its responsible and irresponsible use.
From Productivity Tool to Professional Responsibility
The appeal of GenAI within legal practice is easy to understand. Lawyers routinely work with large volumes of documents, lengthy judgments and extensive bodies of evidence, making tasks such as document review, chronology preparation, summarisation and preliminary research particularly time-consuming.
While GenAI can make many of these processes considerably faster, speed does not guarantee accuracy. These systems can produce convincing but fabricated authorities, inaccurate propositions of law and subtle ‘factual errors’. [vi] The greater risk is not an obviously absurd answer, but one that appears sufficiently credible to escape detection.
The GPN-AI therefore places particular emphasis on human verification. AI may accelerate legal work, but it does not lower the expected standards a lawyer. Where GenAI contributes to a document, the responsible practitioner must ‘independently verify’ its content, including the accuracy and provability of facts, the existence and relevance of authorities, the admissibility of evidence and compliance with procedural and discovery obligations.[vii]
Competent AI use consequently requires more than technological literacy. It requires the professional judgment to recognise when an output warrants further research, scepticism or rejection. AI cannot become a defence to professional error.[viii]
Evidence is Where the Line Becomes Clearer
The distinction between assisting legal work and generating evidence becomes particularly important when AI enters the evidentiary process.
A lawyer’s submissions represent their legal argument. Evidence such as affidavits, witness statements and expert reports, however, purport to represent the recollection, knowledge, experience or professional opinion of another person. AI can assist with preparing these materials, but it cannot supply the authenticity that a human provides.
The GPN-AI reflects this distinction by imposing particular requirements on the use of AI in preparing evidence.[ix] Where AI has been used in a way that ‘informs evidence or opinion’, disclosure may be required.[x] Witness statements and affidavits must remain grounded in the person’s own recollection, knowledge and experience, while expert reports must represent ‘the expert’s own opinion and reasoning ’. [xi]
The integrity of evidence depends not only on the words appearing on the page, but on who is responsible for those words and how they come to exist. The Court does not treat the involvement of AI as automatically undermining reliability. Instead, its approach asks whether the integrity of evidence can be preserved through human responsibility, verification and transparency.
However, this approach is not universal, and the Supreme Court of New South Wales has adopted a more restrictive approach under Practice Note SC Gen 23, particularly in relation to the use of GenAI in generating the content of affidavits, witness statements and other material intended to reflect a person’s evidence or opinion. [xii]
The divergence highlights that Australian Courts are still negotiating the boundaries of AI in legal practice. The question is therefore becoming less about whether lawyers can use AI and more about whether it can be used for particular tasks without compromising the integrity of the legal process.
Privilege Problem
Accuracy is only one side of the problem. Lawyers are more concerned with what happens to confidential information once it enters an AI system.
Legal professional privilege depends upon confidentiality, and disclosing privileged material to a third party in circumstances that are inconsistent with maintaining that confidentiality may place the privilege at risk.[xiii] A lawyer may upload a client’s instructions, a draft deed, or an affidavit to an AI platform for summarisation or analysis, but the convenience must be weighed against how the platform stores, processes and uses the information.[xiv]
The Federal Court’s GPN-AI therefore cautions against entering confidential, privileged or protected material into GenAI systems where doing so could compromise legal or confidentiality obligations. The key question is whether the lawyer remains in control of the information once it enters the system. Simply describing the platform as “secure” does not answer that question.
Closed, firm-hosted, and domain-specific AI systems may provide greater control over sensitive information, but they do not eliminate the risk completely. Law firms must still understand where information is stored, who can access it and whether inputs are retained or used for other purposes.[xv]
This makes AI governance an organisational responsibility, not one that can be left solely to individual lawyers. Firms adopting AI should establish clear policies governing approved tools, the information that may be entered and the work requiring additional review or approval.
The Future Lawyer
The rise of AI is often framed by whether it will replace lawyers. The more immediate question may be whether lawyers will need to adapt. Working alongside AI will require technological literacy, but that alone is not enough. The ability to question AI-generated answers may be more valuable than the ability to generate one. As AI becomes increasingly capable of producing convincing legal language, the ability to question, verify and critically assess its output may become more valuable than the ability to generate it.
The Federal Court’s GPN-AI reflects this shift. Rather than restricting innovation, it establishes a framework in which AI can assist legal practice while humans remain accountable. Its emphasis on accuracy, transparency, confidentiality, human oversight and professional responsibility also provides principles capable of adapting as technology evolves, rather than rules tied to particular AI tools.
Therefore, the future lawyer may not use AI most extensively, but the one who knows when to use it, when to question it and when not to use it at all.
──────────────────────────────────────────
References
[i] Federal Court of Australia, General Practice Note: Use of Generative Artificial Intelligence Practice Note (GPN-AI), 16 April 2026 (‘GPN-AI’).
[ii] Ibid.
[iii] Catie Moore and Lauren Separovich, ‘Federal Court Releases Use of Generative AI Practice Note: key guidance for Using AI in proceedings’ Hall & Wilcox (Web Page, 17 April 2026) <https://hallandwilcox.com.au/news/federal-court-releases-use-of-generative-ai-practice-note-key-guidance-for-using-ai-in-proceedings/>.
[iv] \ GPN-AI (n 1).
[v] Carly Stebbing, ‘Federal Court Embraces GenAI, what does it mean for legal practitioners?’ LEAP Legal Software (Web Page, 13 May 2026) <https://www.leaplegalsoftware.com/au/blog/federal-court-embraces-genai-legal-practitioners/>.
[vi] GPN-AI (n 1).
[vii] Tim O’Callaghan, Travis Shueard and Sahar Al-Dulaimi, ‘The Federal Court’s New Practice Note on the Use of Generative Artificial Intelligence’Piper Alderman (Web Page, 21 April 2026) <https://piperalderman.com.au/insight/the-federal-courts-new-practice-note-on-the-use-of-generative-artificial-intelligence/>.
[viii] Naomi Griffin and Philippa Munton, ‘Using AI in the Federal Court of Australia: A Guide to the New GPN-AI Practice Note’ Clifford Chance (Web Page, 17 April 2026) <https://www.cliffordchance.com/insights/resources/blogs/group-litigation-and-class-actions/2026/04/using-ai-in-the-federal-court-of-australia-a-guide-to-the-new-gpn-ai-practice-note.html>.
[ix] GPN-AI (n 1).
[x] Griffin and Munton (n 8).
[xi] O’Callaghan,Shueard and Al-Dulaimi (n 7).
[xii] Myrna Taouil, ‘The Federal Court Sets the Rules on AI’ Mills Oakley (Web Page, 26 May 2026) <https://www.millsoakley.com.au/insights/the-federal-court-sets-the-rules-on-ai/>.
[xiii] Penelope Ford and Lauren Separovich, ‘Managing legal privilege risks in the age of generative AI’ Business News (Web Page, 5 June 2026) <https://www.businessnews.com.au/article/Managing-legal-privilege-risks-in-the-age-of-generative-AI>.
[xiv] Catie Moore and Lauren Separovich, ‘When Using AI Risks Waiving Legal Professional Privilege’ Lawyers Weekly (Web Page, 19 May 2026) <https://www.lawyersweekly.com.au/biglaw/44317-when-using-ai-risks-waiving-legal-professional-privilege>.
[xv] Ibid.