In November 2025, the Supreme Court of Western Australia released its Guidelines for the Use of Generative Artificial Intelligence. It joins courts around the country—in particular the Federal Court, which released its own practitioner protocol on generative AI, GPN-AI, in April 2026 – in establishing terms of use (and non-use) for generative AI in litigation. One thing stands out about this guidance.
The Court states that “Generative AI does not think or reason. Rather, it predicts word sequences based on training data.” It further cautions that “AI generated content should never be presumed to be correct and is not a substitute for legal research or human experience and oversight.” There is more than meets the eye in this statement.
Beyond calling out fabricated citations (or “hallucinations”), the risks are no longer hypothetical. In Mata v. Avianca, Inc., 678 F Supp 3d 443 (2023), lawyers were sanctioned after submitting material containing non-existent cases generated by ChatGPT. The deeper point is structural: generative AI produces outputs through computational processes fundamentally different from human professional judgement.
Yet as those outputs become increasingly capable of performing tasks associated with legal analysis, the question becomes not simply whether lawyers can verify them, but what forms of professional judgement must remain with the human practitioner.
From assistant to participant
Lawyers are not strangers to legal technology. Electronic research databases and document management systems have been staples in Western Australian legal practices for years. What is different about generative AI is that it does not just fetch you information – earlier tools still required you to read, digest and apply the information they retrieved. Generative AI drafts. It summarises. It analyses: it gives you something that looks an awful lot like the first cut at an answer, rather than a pile of working materials.
This shift is already visible in legal education, not just practice. Bond University’s law school has woven “Protégé”, LexisNexis’ generative AI research and drafting tool, into their teaching. Flinders University Law School also plans to make that tool available for students this semester. And the University of Queensland’s TC Beirne School of Law was recently the subject of a case study on how best to adapt legal academia to teach these new AI tools. Taken together, these developments indicate a wider shift towards incorporating AI into Australian legal education.
The obligation the guidelines assume
The Guidelines, similar to guidance in other jurisdictions, confirm that verification is the responsibility of the practitioner. AI-generated information should always be checked against primary authorities. Sensitive or privileged information should never be input into publicly available AI platforms. All of these rules make sense. They are prudent. But they are predicated on another idea worth underscoring: the practitioner checking AI-generated output already knows enough substance and exercises enough professional judgement to know when something looks wrong.
If a lawyer cannot recognise an invented case citation or statutory reference that does not apply to the jurisdiction at issue, they cannot fulfil their duty to verify. Meeting the requirement becomes nearly impossible.
That capacity to recognise depends on something it does not inherently confer: traditionally, at least in part, knowledge of substance and judgement is learned by doing the very tasks generative AI will come to perform with ease. Researching legal authorities from scratch. Drafting the first iteration of a legal memo. Line-by-line editing of a document.
According to LawCPD’s 2026 report on legal industry trends, one-third of Australian lawyers surveyed had never used AI. Recognising this gap, the Legal Practice Board of Western Australia joined both the Law Society of New South Wales and the Victorian Legal Services Board and Commissioner in signing a joint statement on the responsible use of AI. Regulators from each jurisdiction that relies on the Uniform Law agreed to revisit and update their guidance as technology advances.
Legal education, at every level, faces the same question
That is where I would argue the conversation needs to extend. University law schools have already started to think about how they will teach research, writing, and analysis when artificial intelligence can produce credible first drafts of all three. But the same structural question also applies to vocational and competency-based legal education, one of the pathways through which legal assistants, legal secretaries and legal support staff are trained.
Here in Western Australia, vocational legal education, delivered through TAFE and other registered training organisations, is built around nationally recognised units of competency. These units are made up of assessable skills or capabilities (conducting legal research, drafting correspondence, using professional technology tools, and so on) which have defined sets of requirements that a student must be able to meet to qualify as “competent” in that skill.
But if a student were to complete a research or writing task with significant assistance from AI, what have they actually shown competence in: legal research, or the operation of a tool which can do that research for them? Two different skills; a competency-based training model needs to be clear about which one it is testing.
These questions also extend beyond the classroom. Legal support staff undertake research, drafting, document management and other tasks that increasingly intersect with AI-enabled legal practice. Preparing legal support staff to work responsibly with AI has implications for their training, but also for the law firms and lawyers who will rely on them.
What this means going forward
As for lawyers presently practising in Western Australia, the Supreme Court provides clear guidance about what is expected of them today: verify information before relying on it, protect confidential information, and remember that a fast answer is not necessarily a correct answer for the jurisdiction and facts at hand.
When it comes to legal training — both in law schools and on the job — a far more open-ended question exists: where will tomorrow’s practitioners and paraprofessionals learn the judgement necessary to meet these obligations at exactly the point when some of the tasks that traditionally taught that judgement are now being assigned to the tools they must learn to verify?
There is no single answer. How we arrive at one is a conversation worth pursuing collectively across the state’s legal profession, law schools and vocational education providers, rather than addressing the challenge in isolation.
