AI document review and Australian courts: what legal professionals need to know

Can I stand behind the results?

The answer is 'yes' if you understand what the courts expect and choose your tools carefully. 

What Australian courts are saying about AI 

Australian courts have moved quickly to provide guidance on AI use in proceedings, and there is consistency across every jurisdiction that has issued formal guidance. AI is welcome but accountability remains with the lawyer and the type of AI tool you use matters significantly.  

The Federal Court of Australia issued its Use of Generative Artificial Intelligence Practice Note (GPN-AI) in April 2026. The Court recognises that Generative AI has the potential to facilitate the just resolution of disputes by increasing efficiency and reducing legal costs, and makes it clear that its use must be responsible and consistent with existing professional obligations.   

In NSW, Practice Note SC Gen 23 commenced on 3 February 2025, setting out obligations on practitioners and witnesses where AI has been used in preparing documents for court, including expert reports. 

In Victoria, Practice Note SC Gen 25 was issued on 14 May 2026 and replaced the earlier 2024 guidelines and established that court users are responsible for the content of their court documents whether or not they have used AI. Importantly for document review practitioners, the Victorian guidelines acknowledge the role that AI, in the form of Technology Assisted Review (TAR), already plays in reducing the time and cost of large-scale document review, and express the view that specialised, legally focused AI tools are likely to be more useful and reliable for parties in litigation than general purpose AI tools. This is a significant and direct endorsement of purpose-built legal AI tools from the court itself.  

In Western Australia, the Supreme Court issued a consultation note on a proposed AI practice direction and invited submissions from the profession in 2025. The WA Law Society’s response advocated for a balanced, principles-based approach that enables legal practitioners and court users to take advantage of Generative AI while ensuring the integrity of evidence, submissions and other documents. A formal practice direction is expected to follow. In the meantime, WA practitioners are subject to the same professional obligations that apply nationally, maintaining client confidentiality, providing independent advice and verifying the accuracy of any AI-assisted work before it is filed or relied upon. 

In South Australia, the Judges of the Supreme Court approved Guidelines for the use of Generative AI following consultation with the profession throughout 2025, adopted from 1 January 2026. The guidelines apply across the Supreme Court, District Court, Magistrates Court, Youth Court and Environment, Resources and Development Court, giving practical examples of ethical and responsible AI use alongside examples of improper use. The Court envisages that AI, when used correctly, has the potential to change the way court cases are heard and determined, leading to cost savings and making courts more accessible. 

In Queensland, Practice Directions issued in 2025 specifically target AI hallucinations in submissions and require experts to disclose AI use in criminal proceedings. This is in response to real cases before Queensland courts and tribunals where self-represented litigants were warned against relying on AI hallucinations.  

The consequences of misuse have already been severe. In August 2025, a Victorian solicitor lost the ability to practise as a principal lawyer, handle trust money and operate their own practice after submitting fictional AI-generated cases to the Federal Circuit and Family Court of Australia. 

Three questions every lawyer should ask 

Across every jurisdiction, the courts’ expectations distil into three core requirements that map directly to any AI-assisted document review workflow. 

1. Can you explain how the AI reached its conclusions? All persons are expected to be able to inform the Court as to what Generative AI was used, how it was used and for what purpose. This means ‘the AI found it’ is not a sufficient answer. The tool you use must show its reasoning in a way that a supervising lawyer can audit, verify and explain. 

2. Have you confirmed the results are accurate? The GPN-AI is explicit that AI tools can produce results that are not accurate, entirely fictitious or plainly wrong. Where AI has been used to prepare documents filed with the court, the responsible lawyer must have confirmed the facts are based on what can be proved, legal authorities exist and support the propositions stated, and chronologies are accurate. The obligation to verify does not disappear because a machine did the initial work. 

3. Have you protected confidential and privileged information? The risks of inadvertent disclosure may be lower for tools that operate in closed or controlled environments. This is a critical distinction for document review. Tools that operate within your existing secure review environment present a fundamentally different risk profile to open AI tools that may store, process or learn from data you upload. 

Why the type of AI matters 

Not all AI tools are equal from a legal risk perspective, and the courts’ consistent distinction between open and closed AI environments is directly relevant to how eDiscovery teams should approach AI-assisted review. 

Open AI tools, where data is transmitted to external servers, potentially stored or used for model training, carry significant risk in a legal context. Documents under compulsory process, privileged communications and confidential client information should never enter an environment where you cannot be certain of what happens to them. 

Closed AI tools that operate within a controlled, ringfenced environment using your existing data present a far lower risk profile, provided the tool can show its reasoning and produce auditable and explainable results.

This is exactly why eDiscovery Collab selected Claira for AI-assisted document review. Claira operates entirely within the Nuix Neo Discover environment so documents never leave the secure review platform that your team already uses. Every result includes structured reasoning that supervising lawyers can read, verify and rely on. Your document content is never used to train or fine-tune AI models. 

Defensible review in practice

Defensible AI-assisted review is not simply a technology question. It requires the right tool used in the right way, with experienced professionals overseeing the process. 

At eDiscovery Collab, our senior-led team works with you throughout every Claira matter to help design prompts, test results against your review criteria, apply QC processes to validate AI outputs against human coding and ensure the review decisions your team certifies are traceable and defensible. The AI accelerates the work but the lawyers remain accountable for it. 

The courts are not asking legal professionals to avoid AI but they are asking them to use it responsibly. With the right tools and the right team alongside you, that is entirely achievable. 

Click here to find out how Claira supports defensible AI review.

This blog is general information only and practitioners should consult the relevant practice notes and their professional obligations directly.