The Dayal Case: What Every Australian Lawyer Needs to Know About AI Generated Citations
Since the 2025 Dayal case, over 20 Australian matters have involved AI fabricated citations. Here is what your firm needs to do about it.
In August 2025, a Victorian legal practitioner was sanctioned for submitting AI generated citations that did not exist. The Dayal case was the first time an Australian court formally addressed the problem of fabricated AI references, but it was far from the last.
Since then, more than 20 Australian cases have involved AI generated false references, according to tracking by Barry Nilsson. Globally, the rate has reached 4 to 5 documented new cases per day, based on data compiled by the National Law Review. This is not a theoretical risk. It is happening now, across jurisdictions and practice areas.
What happened in Dayal
The facts are straightforward. A practitioner used an AI tool to assist with legal research and submissions. The AI produced citations that looked legitimate (correct formatting, plausible case names, realistic citation numbers) but referred to cases that did not exist. These fabricated citations were included in court filings without independent verification.
The court took a dim view. The practitioner faced professional sanctions, and the case became a reference point for every subsequent discussion about AI use in Australian legal practice.
Why this keeps happening
AI language models, including ChatGPT, Copilot, and legal specific tools, generate text by predicting what words are most likely to come next. When asked for a citation, the model produces something that looks like a citation because it has seen thousands of real ones during training. But it is not searching a legal database. It is generating text, and sometimes that text refers to cases that do not exist.
This is not a bug that will be fixed in the next software update. It is a fundamental characteristic of how these models work. The technical term is "hallucination," and while the rate has improved over time, it has not been eliminated. Even tools marketed specifically for legal research can produce plausible but incorrect references.
The common assumption that legal specific AI tools are immune to this problem is wrong. These tools reduce the rate of fabrication compared to general purpose models, but they do not eliminate it. Any AI output that will be filed with a court or sent to a client needs independent verification.
How courts are responding
Australian courts have moved quickly. Victorian, NSW, and Federal courts now require practitioners to independently verify any AI generated output before filing. Several courts have introduced practice notes or directions specifically addressing AI use.
The obligations are not complicated, but they are explicit. If you used AI to assist with research, drafting, or analysis, you are expected to verify the output against primary sources. In some jurisdictions, you may be required to disclose that AI was used. Ignorance of these requirements is not a defence. They have been widely publicised since mid 2025.
The professional conduct implications are significant. Submitting fabricated citations engages duties of honesty and candour to the court, competence obligations, and potentially the duty to maintain trust and confidence in the profession. A single incident can result in costs orders, referral to the relevant law society, or disciplinary proceedings.
What your AI usage policy needs to cover
If your firm does not have a written AI usage policy, you are operating without a safety net. If you have one but it does not specifically address citation verification, it has a gap that needs closing.
An effective policy covers five areas.
Permitted use cases define where AI tools can and cannot be used. Research assistance and first draft generation are reasonable use cases. Final drafting of court submissions without human review is not.
Verification requirements are the most important element. Every citation, case reference, legislative reference, and factual claim generated by AI must be verified against a primary source before it is included in any document that leaves the firm. This applies to internal memos that might later be relied on, not just court filings.
Disclosure obligations need to reflect the requirements of the jurisdictions your firm practises in. Some courts require disclosure of AI use. Even where not mandated, firms should consider whether disclosure is prudent in specific circumstances.
Training requirements ensure that every person in the firm who has access to AI tools understands the verification obligations. This includes lawyers, paralegals, and administrative staff who may use AI for correspondence or document preparation.
Review and update schedule keeps the policy current. Court requirements and AI tool capabilities are both changing rapidly. A policy written in 2025 may not reflect the obligations and tools of 2026. Quarterly review is a reasonable cadence.
The practical reality
None of this means you should stop using AI. The productivity benefits are real, and firms that refuse to adopt AI tools will fall behind. But the adoption needs to happen within a framework that protects the firm and its clients.
The lawyers who get into trouble are not the ones using AI. They are the ones using AI without verification. The fix is not to ban the technology. It is to build a simple, enforceable check into every workflow where AI generated content could end up in a court filing or client communication.
Next step
If your firm does not have an AI usage policy that covers citation verification, we can help you build one. ServiceScaler works with Australian law firms to develop practical AI policies that reflect current court requirements and fit into existing workflows. The goal is a policy your team will actually follow, not a document that sits in a drawer. Get in touch to discuss what your firm needs.