Judge describes explanation for making admission as plausible
In a proceeding arising from an applicant’s defamation claim against a respondent, Australia’s Federal Court disagreed with an argument that the applicant’s explanation for making a certain admission was nonsensical or characteristic of an artificial intelligence (AI) hallucination.
Finnegan v Kemner (No 3) [2026] FCA 1402 arose from an email the applicant had sent to multiple recipients on 8 April 2019.
The applicant claimed that the respondent had defamed him. One of the allegedly defamatory matters was the respondent’s online noticeboard post stating that the applicant had been “trying to be anonymous and failing” when sending the email.
In paragraph 4(c)(i) of a substituted defence filed on 20 November 2025, the respondent appeared to allege a defence of justification by pleading that the applicant had anonymously sent the email using an external affairs email address.
In paragraph 4(c)(i) of a substituted reply filed on 5 December 2025, the applicant responded to the allegation by simply admitting to it.
On 21 September 2026, the first trial day, the applicant applied for leave under r 16.53 of the Federal Court Rules 2011 (Cth) to amend paragraph 4(c)(i) of his substituted reply to withdraw the admission.
Through the proposed amendment, the applicant sought to admit that the email did not identify him by name, but deny that he sent it anonymously, given that the email identified the sender as “External Affairs,” a business name he was using.
In opposing the leave application, the respondent alleged that the applicant’s explanation for making the admission made no sense and was a verbose hallucination characteristic of the AI that the applicant utilised to produce almost all his recent communications for this matter.
The Federal Court granted the applicant leave to amend paragraph 4(c)(i) of his substituted reply. The court directed him to file and serve an amended substituted reply, limited in line with the court’s orders.
The court also ordered the applicant to pay the respondent’s application costs and any costs thrown away due to the reply’s amendment and/or the admission’s withdrawal.
The Federal Court ruled that granting leave would serve the interests of the administration of justice.
The court added that allowing a modest change would not risk a loss of public confidence in the legal system, would not be unjust to the respondent, and would enable the determination of the real matters in controversy.
The court attributed the admission to inadvertence. The court noted that the question of whether the applicant sent the email anonymously might be an issue in the proceeding.
Regardless of the lateness of the withdrawal and the amendment, the Federal Court saw minimal prejudice to the respondent’s conduct of the case.
The court pointed out that the respondent’s counsel mentioned needing to change the case strategy, the opening and closing submissions, and the cross-examination approach.
The court explained that the amendment would align with the applicant’s evidence-in-chief and would only impact a dispute regarding the characterisation of the facts concerning the email.
As the applicant had asserted, the court noted that the parties had agreed to that characterisation in a joint statement of the legal and factual issues filed on 28 July 2026 and in other portions of the pleadings.
The Federal Court rejected the respondent’s argument that the applicant’s explanation for the admission made no sense or showed signs of an AI hallucination.
The court acknowledged that the applicant’s explanation could have been fuller, especially in connection with when he had discovered the mistake. However, the court found the explanation plausible, sufficient in the circumstances, and reflective of the withdrawal’s relatively minimal nature.
Here are some other recent decisions by the Federal Court of Australia.
On 14 September 2026, the court forwarded some material filed in a superannuation proceeding to the Legal Professional Board of Tasmania and the Queensland Legal Services Commission for a determination of whether they should further investigate the applicant’s lawyer.
In an insurance matter, the court issued a 28 August 2026 decision referring to a proposed pleading as “AI slop” and said it fell short of its expectations in [1.3], [1.4], and [2.1] in its Use of Generative Artificial Intelligence Practice Note (GPN-AI).
Enjoy this story? Read the latest litigation and dispute resolution news on the main page!
