AI tools have evolved into useful assistants in the legal profession – when used properly. They can sort and compare documents, summarise a brief, comb through multiple Acts and regulations for relevant sections, and find potentially relevant authorities in seconds.

However, in litigation, where accuracy is everything, the courts have made it clear that AI should only be used as a tool to assist. The lawyer remains accountable for checking and verifying the information produced.

In a nutshell: ‘trust but verify’.

This article explains where Australian courts currently stand on the use of generative AI in litigation, why it matters, and how our approach compares with the rest of the world.

 

Why This Matters Now

Over the past two years, Australian courts have moved quickly to regulate how AI can be used in legal proceedings.

One of the concerns prompting this response is the risk of lawyers using AI tools without properly understanding their limitations. Vague or generic prompts, rather than detailed and context-rich instructions, can increase the risk of inaccurate or misleading outputs.

The less precise the prompt, the greater the risk that AI will generate inaccurate information or ‘hallucinate’ facts, cases or legal principles. The problem becomes much more serious when that information is relied upon without being independently verified through proper legal research.

These hallucinations are not a hypothetical risk.

In 2024, a Victorian solicitor became the first Australian lawyer formally penalised over AI misuse after filing a list of false cases as legal authorities in court. The fictitious cases were generated using AI legal software and were not independently verified by the solicitor. When the Federal Circuit and Family Court attempted to verify the citations, the cases could not be found.

Another incident occurred in Western Australia, where a lawyer was referred to the legal regulator after court documents were filed containing AI-generated case citations that did not exist.

Similarly, during a murder trial in the Supreme Court of Victoria, defence lawyers submitted documents containing non-existent case judgments and fabricated quotations attributed to a speech to the state legislature.

These cases demonstrate why verification is so important. AI can produce information that appears authoritative and convincing, even when it is wrong.

 

How Australia Compares with the Rest of the World

Australia is far from alone.

Since the now-infamous US case of Mata v Avianca in 2023, where lawyers were sanctioned after submitting AI-generated fictitious cases, courts around the world – from Europe to Asia – have grappled with the same problem.

Australia has responded relatively quickly to the global AI boom, with courts issuing practice notes and guidelines addressing its use in legal proceedings.

At the same time, AI adoption within the legal profession continues to grow. Many legal professionals now use AI as part of their daily workflow, but the challenge is ensuring that its use is supported by appropriate policies, supervision and safeguards.

 

The Rules Across Australian Courts

In Australia, individual courts have issued their own practice notes and guidelines. Key examples include:

  • NSW Supreme Court: Practice Note SC Gen 23 – Use of Generative Artificial Intelligence (AI), which has been in force since early 2025 and is periodically reviewed.
  • Federal Court of Australia: The Generative AI Practice Note (GPN-AI).
  • Supreme Court of Victoria: Practice Note SC Gen 25 and accompanying guidelines on the use of artificial intelligence by court users.

While the requirements differ between courts and the type of document involved, some common principles are emerging.

Evidence must remain the deponent’s own

Particular care must be taken when AI is used in connection with affidavits, witness statements and other evidentiary material.

Evidence must genuinely reflect the deponent’s own knowledge, recollection and evidence. Lawyers and clients should therefore understand the relevant court requirements before using generative AI to assist with documents intended to be filed or relied upon in proceedings.

Verify everything

Courts expect lawyers to personally check that every case, statute and quotation cited actually exists, says what it is claimed to say and applies in the relevant jurisdiction.

Relying on AI to ‘verify’ its own output is not enough.

Lawyers also need to remain mindful of both AI error and human error when developing prompts, interpreting results and using AI-generated material as part of legal research.

Protect confidential information

Suppressed, non-public, confidential or privileged material should not be entered into public AI platforms without careful consideration of how that information will be handled.

Depending on the platform, information entered as prompts may be stored, processed or used in ways that create confidentiality and privacy risks.

Existing obligations concerning confidentiality, legal professional privilege and suppression orders continue to apply regardless of the technology being used.

Be ready to disclose

Depending on the court and the circumstances, lawyers may need to disclose when and how AI has been used.

If questioned, a lawyer should be in a position to explain which AI tool was used, the extent to which it was relied upon and the steps taken to check and verify its output.

 

Practical Takeaways for Clients

If your matter involves litigation, there are a few things worth keeping in mind:

  • AI can assist, but your lawyer remains accountable. Technology does not replace a lawyer’s professional obligations or responsibility for work filed with the court.
  • Your evidence must remain yours. Affidavits and witness statements need to accurately reflect your own knowledge, recollection and evidence.
  • Ask how your law firm uses AI. A good firm should be transparent about where AI can assist with efficiency, research or document review – and where professional judgement and verification remain essential.
  • Confidentiality is non-negotiable. Sensitive details about your matter need to be handled carefully and should not simply be entered into public AI tools.
  • Efficiency should not come at the cost of accuracy. Faster is only better if it is also right.

 

The Bottom Line

Used well, AI can be a genuine asset in litigation. It can help reduce costs, speed up research and document review, and give lawyers more time to focus on strategy, analysis and advocacy.

Used carelessly, however, AI can produce inaccurate or entirely fabricated information. If that information is relied upon without proper verification, the consequences can be serious – for the client, the lawyer and the administration of justice.

The Australian courts have drawn the line clearly: technology may assist, but lawyers remain accountable for the work they produce.

When it comes to AI in litigation, the principle is simple: trust, but verify.