Artificial intelligence has moved from a novelty in Australian legal practice to something courts are now actively writing rules about. It is not deciding cases, no Australian court hands a verdict to an algorithm, but it is already involved in how evidence is prepared, how research is done, and increasingly, how courts expect lawyers to behave when they use it.
This guide explains what AI is genuinely being used for in Australian legal practice right now, what courts have started requiring in response, and a real, cautionary example of what can go wrong when it is used carelessly.
What AI Is Actually Doing in Australian Legal Practice
Lawyers increasingly use AI tools for legal research, summarising large volumes of documents, and assisting with early drafting of routine material.
In large scale disclosure exercises, sometimes called e-discovery, AI-assisted document review and classification tools help legal teams sort through huge volumes of material far faster than manual review alone.
None of this amounts to AI making legal decisions. Every documented use in Australia is assistive, supporting lawyers and court staff rather than replacing the judgement of a judge, tribunal member, or admitted legal practitioner.
It is more accurate to describe AI as already shaping how cases are prepared than as deciding how they are resolved.
This distinction matters because headlines about AI in the legal system can easily overstate what is actually happening.
A tool that helps a lawyer find relevant case law faster, or that flags a clause worth reviewing in a large contract, is meaningfully different from a system that determines the outcome of a dispute. Australian courts have been careful to keep that line clear in the guidance they have issued so far.
Courts Are Now Writing Rules for This
Several Australian courts have moved quickly to formalise how AI can and cannot be used in litigation. The Federal Court of Australia issued its Use of Generative Artificial Intelligence practice note, setting out guidance for parties, lawyers, and litigants on when and how generative AI can be used in preparing court documents and submissions.
The NSW Supreme Court went further, issuing a dedicated Generative AI Practice Note and Judicial Guidelines that restricts AI use for drafting affidavits, witness statements, and expert reports, among other safeguards for solicitors and barristers appearing before the court.
Other Australian courts and tribunals have introduced or are developing similar guidance, and the picture is best described as a fast moving, court by court patchwork rather than one settled national rule, since each court has taken a slightly different approach to what it will and will not allow.
This patchwork approach can be confusing for anyone outside the profession, but the underlying theme across every practice note issued so far is consistent.
Courts are not banning AI outright, but they are insisting that any output generated with its help be treated as a first draft requiring genuine human verification, not a finished, reliable product on its own.
The Cautionary Example: Fabricated Citations
The clearest illustration of why these rules exist happened in 2025, when a Victorian lawyer was reported to have filed court documents in a murder case containing AI-fabricated quotes and citations to judgments that did not actually exist.
This kind of error is often called an AI hallucination, where a generative AI tool produces confident, plausible-sounding output that turns out to be entirely fabricated.
The case was widely reported internationally and is generally described as one of the first Australian instances of a lawyer being penalised for this kind of AI-related error in court filings.
It prompted renewed attention from courts and professional bodies on verification obligations, reinforcing that anything generated with AI assistance still needs to be independently checked before it goes anywhere near a court document.
How the Profession Is Responding
State legal bodies have moved to provide practical guidance for solicitors navigating this shift. The Law Society of NSW’s guide to the responsible use of artificial intelligence sets out expectations for how solicitors should verify AI-assisted work, maintain client confidentiality when using AI tools, and disclose AI use where relevant.
The Law Society also maintains a running tracker of court protocols on AI use across Australia, reflecting how quickly this area continues to change as more courts and tribunals introduce their own specific rules.
A Word on Automated Decision Making More Broadly
It is worth distinguishing generative AI in litigation from a separate, older concern in Australian legal and policy circles about automated government decision making.
The Robodebt scheme, an automated debt recovery program run by Services Australia that was ultimately found to be unlawful and became the subject of a Royal Commission, is the case most often cited as a warning about accountability gaps when automated systems make consequential decisions without adequate human oversight.
Robodebt predates generative AI and was not a court process, but it remains a frequently referenced example in legal commentary about why human oversight of any automated or AI-assisted decision matters, particularly where the outcome significantly affects someone’s life.
What This Actually Means If You Are Involved in a Case
For someone going through a legal matter, none of this needs to be alarming, but it is worth understanding what reasonable AI use looks like from your own lawyer, and what questions are fair to ask.
- It is reasonable to ask your lawyer whether AI tools were used in preparing documents relevant to your matter, particularly for anything that will be filed with a court.
- A competent lawyer should be able to explain how any AI-assisted research or drafting was independently checked before being relied upon.
- You should expect any case citations, quotes, or factual claims in your documents to be verifiable against real sources, regardless of how they were originally drafted.
- If your matter involves large volumes of documents, such as in a commercial dispute, AI-assisted review tools may genuinely speed up the process and reduce costs, which can work in your favour.
- Ultimately, professional responsibility for the accuracy of anything filed in your case remains with your lawyer, not with any tool they used to help prepare it.
Conclusion
AI is genuinely reshaping how Australian legal cases get prepared, from research to document review, but it is not deciding outcomes, and courts have moved quickly to put guardrails around how it can be used.
The 2025 fabricated citation case is a clear reminder that AI output still needs careful human verification before it goes anywhere near a courtroom.
If your matter involves questions about how evidence or documents were prepared, lawyer.com.au can help you find a lawyer to guide you through it.
FAQs
1. Is AI actually deciding legal cases in Australia?
No. There is no evidence of AI making legal decisions in Australian courts. Every documented use is assistive, such as research or document review, with judges and lawyers remaining the decision makers.
2. Do Australian courts have rules about using AI in litigation?
Yes, several courts, including the Federal Court and the NSW Supreme Court, have issued formal practice notes governing how generative AI can be used in preparing court documents and submissions.
3. What happened with the AI fabricated citations case in Australia?
In 2025, a Victorian lawyer filed court documents in a murder case containing AI-generated quotes and case citations that did not exist, prompting renewed scrutiny of AI verification obligations for lawyers.
4. Can lawyers use AI to draft legal documents?
Generally yes for some purposes, but courts and law societies increasingly require disclosure and independent verification, and some courts restrict AI use for specific documents like affidavits and expert reports.
5. What is Robodebt and how does it relate to AI in law?
Robodebt was an unlawful automated government debt recovery scheme that predates generative AI. It is frequently cited in legal commentary as a warning about the risks of inadequate human oversight over automated decision systems.