What happened
It was reported that in early 2025, a junior solicitor at Melbourne firm Massar Briggs Law was preparing footnotes for an amended native title application involving the Wamba Wemba people. The references were important because they identified the historical, anthropological and legal material supporting the claim group’s connection to Country.
The solicitor was working from home and could not access the firm’s physical library or the electronic documents held on its internal network. Instead, she turned to online research tools to prepare the citations. The court later found that some references were fabricated, while others referred to genuine documents but contained incorrect information. The judge considered it likely that generative AI had been involved, possibly through Google Scholar, although the precise tool was not established.
The problems came to light when First Nations Legal and Research Services attempted to locate the material cited in the application. Many of the references could not be found, while others did not support the statements attributed to them. What should have been a routine exercise in checking supporting documents became a substantial task of identifying and correcting unreliable references.
Massar Briggs Law acknowledged what had happened. The court required the applicants’ solicitors to file affidavit evidence, including from the junior solicitor, explaining the circumstances, among them her lack of access to the source material and the absence of adequate checking before the documents were filed.
In Murray on behalf of the Wamba Wemba Native Title Claim Group v State of Victoria [2025] FCA 731, the Federal Court ordered the firm to pay the respondents’ costs on an indemnity basis. This is a more substantial costs order than the usual arrangement and can require payment of most of the reasonable costs incurred by the other parties.
Importantly, the order was made against the solicitors rather than the Wamba Wemba claim group. The court recognised that the applicants themselves were not responsible for the errors and should not have to bear the financial consequences. The judge also decided not to refer the solicitors to the Victorian Legal Services Board.
What an auditable version would have shown
The court’s findings pointed to a basic problem in the preparation of the application. References had been included without anyone adequately checking them against the original material.
An auditable process would have shown which sources the junior solicitor used, which documents she could access and whether the citations had been independently verified. It would also have made clear to the supervising solicitor that some references had been prepared without access to the firm’s underlying research materials.
A CitationVerifier could have helped identify references that could not be matched to genuine judgments, reports or other documents. Native title research, however, often relies on historical and anthropological material that may not be available through standard legal databases. Those sources would still need to be checked manually against the originals.
A ConductRecord could have preserved the verification results and the supervising solicitor’s approval before the application was filed. The public decision establishes that adequate checking did not occur, but it does not establish every detail of the firm’s internal record-keeping.
Where the gap was
The most important finding in this case concerned supervision.
The judge recognised that the junior solicitor was inexperienced and working without access to the materials she needed. But the responsibility did not end with her. The firm had failed to put adequate systems in place to ensure that her work was checked before it reached the court.
Headlights identifies a rules-behaviour gap between the professional obligation to verify court documents and the process actually followed. The references were supposed to support a legal claim, yet they were submitted without adequate confirmation that the underlying material existed or said what the footnotes claimed.
The incident also raises the possibility of automation bias, where information produced by a research tool is accepted without sufficient independent scrutiny. The court considered generative AI a likely source of the errors, but did not conclusively identify the tool. It would therefore be misleading to attribute every incorrect citation to a particular AI system.
There is a wider workplace issue here. The solicitor was expected to complete research from home without access to the firm’s source library. Using an online tool may have appeared to be a practical way around that problem, but it did not remove the need to check the original documents.
The failure was not simply that a junior employee made mistakes. It was that those mistakes passed through the firm’s review process and reached the court. The respondents then had to spend time investigating references that should have been verified before filing.
These behavioural classifications are Headlights’ analysis of the reported conduct, not additional findings made by the court.
What governance should have looked like
The first safeguard should have been access to the materials required for the work. If a solicitor cannot examine the original sources, the firm needs to provide access or arrange for the references to be checked by someone who can.
The second is meaningful supervision. Before the amended application was filed, a supervising solicitor should have confirmed that the footnotes referred to genuine documents and accurately represented their contents. That responsibility becomes particularly important when junior staff are preparing material for complex litigation.
A CitationVerifier could assist with references available in recognised databases, while a VerificationGate could require unresolved citations to be reviewed before filing. A ConductRecord could preserve the checks undertaken and identify who approved the final document.
None of these controls would remove the need for professional judgement. A source can exist and still be incorrectly described. In native title proceedings, where historical and anthropological evidence can be central to the claim, checking what a document actually says is just as important as confirming that it exists.
The Federal Court’s decision made the consequences clear. When a law firm’s inadequate checking forces other parties to investigate unreliable material, the firm may be required to pay for the work its mistakes caused.
Failure Pattern: a junior solicitor prepared references for a native title application using online research tools without access to the firm’s original materials. Fabricated and incorrect citations were filed, and the Federal Court ordered the firm to pay the respondents’ costs on an indemnity basis.
Governance Principle: law firms must ensure that research is independently verified and appropriately supervised before court documents are filed. The responsibility remains with the firm, regardless of the technology used or where the work was completed.
The reference implementation of CitationVerifier, VerificationGate and ConductRecord is open source. It lives at github.com/saffronandindia/headlights-oss, Apache 2.0 licensed and free to install. The repository is public now.
Sources
- Murray on behalf of the Wamba Wemba Native Title Claim Group v State of Victoria [2025] FCA 731 (AustLII)
- Solicitor to Pay Costs for Submissions Affected by the Use of AI (Hearsay)
- Melbourne law firm caught using fake AI citations (Information Age, ACS)
- Melbourne firm submits AI-fabricated citations (Lawyers Weekly)
- Australian Federal Court orders costs against law firm over use of Generative AI (Mills Oakley)
- Case alert: A warning on the use of Artificial Intelligence in native title claims (MPS Law)