The lawyer has been referred to Western Australia's legal regulators after using artificial intelligence in preparing court documents for immigration cases. The documentation included AI-creating case citations for cases that were not present.
This is one of more than 20 cases in which AI uses led to warnings from court judges as AI uses resulted in false quotes or other errors, just as they are wary of using technology in legal professions.
In a federal court ruling released this week, the anonymous lawyer was referred to the Western Australian Legal and Judiciary Committee for consideration and was ordered to pay a federal fee of $8,371.30 after filing a submission into an immigration case.
Judge Alain Gerrard said the case “shows an inherent danger associated with practitioners who rely solely on the way in which they interact with the practice's duties.”
In an affidavit, the lawyer told the court that he had relied on the Claude AI of Mankind as a research tool to identify potentially relevant authorities and improve my legal arguments and position, and later validated the submission using a Microsoft co-pilot.
The lawyer said “we developed an overconfidence in relying on AI tools and were unable to properly verify the results produced.”
Sign up: AU Breaking NewsEmail
“There was a false assumption that the content generated by AI tools was inherently reliable, and we now ignore independent verification of all citations through established legal databases,” the lawyer said in the affidavit.
The lawyers apologised to the court and minister's lawyers for the error.
Gerrard said the court “does not adopt a Luddite approach” to the use of generator AI, and that it understood why the complexity of immigration law makes AI tools attractive. However, he warned that there are cases of “number concerns” that led to the AI citing fictional incidents.
Gerrard said the “good cases that are undermined by the incompetence of the rank” and the prevalence of such cases risked “a significant wasting of time and resources on the other side and the court.” He also said it put a risk to the damage to the legal profession.
Gerrard said the lawyers “do not fully understand what he needs,” and not merely confirming that the cases cited were not fake, but also reviewing those cases thoroughly was not enough.
“Legal principles are more than just slogans that can be pasted to submit without context or analysis.”
Since the explosion of the generation AI tools in 2023, at least 20 cases of AI hallucinations have been reported in Australian courts.
Last week, a Victorian Supreme Court judge criticized a lawyer acting for a juvenile accused of murder for submitting misleading information to the court after failing to review documents created using AI.
The document contained references to citations from nonexistent cases and inaccurate citations from Congress' speeches.
After the newsletter promotion
There have also been similar cases involving lawyers in New South Wales and Victoria over the past year, and they have been referred to the state regulatory bodies.
The spy in the case is not limited to qualified lawyers. In the NSW Supreme Court decision, the litigator who expressed himself in a trust case recognized by Chief Andrew Bell in this month's Supreme Court decision, used AI to prepare her speech for an appeal hearing.
Bell said in his decision that he has not criticized anyone who said he was doing his best to represent him. However, he said the problem with using AI to prepare for submissions has been exacerbated when the technology was used by non-representative litigators who said it was “not subject to the professional and ethical liability of legal practitioners.”
He said the use of generative AI tools “may introduce additional costs and complexity,” and “adds to the burden of other parties and their corresponding courts.”
“The fact that generative AI could contribute to improving access to justice, the fact that it is clearly a commensurate goal in itself, stating the need for judicial vigilance, particularly by litigators in litigation.”
Juliana Warner, Australia's Legal Council, said sophisticated AI tools provided a unique opportunity to support legal professions in administrative tasks, but reliance on AI tools did not reduce the professional judgment that legal practitioners would expect to bring to client problems.
“If these tools are being used by lawyers, this must be done with great care,” she said. “Attention must always keep in mind the professional and ethical obligations to the court and its client.”
Warner added that while the court is related to cases where AI generates false quotes as “serious concerns,” given the widespread use of generative AI, a widely parsed ban on use in legal proceedings “is neither practical nor proportional and risks impeding access to innovation and justice.”
Quick Guide
Please contact us about this story
show
The best public interest journalism relies on direct accounts from people of knowledge.
If you have anything to share about this subject, please contact us with a secret using the following methods:
Secure Messages in Guardian App
The Guardian app has a tool to send tips about stories. Messages are end-to-end encrypted and implied within the routine activity that all Guardian mobile apps perform. This prevents observers from knowing you are in absolutely communication with us.
If you don't have the Guardian app yet, download it (iOS/Android) and go to the menu. Select Secure Message.
Securedrop, Instant Messenger, Email, Phone, Posting
If you can safely use the TOR network without observing or monitoring it, you can send messages and documents to the Guardian via the SecureDrop platform.
Finally, the Guardian.com/Tips guide lists several ways to safely contact us, explaining the pros and cons of each.
