Federal Court: Barron v BT Funds AI Pleading Denied as 'AI Slop'

• Australia's Federal Court, in Barron v BT Funds Management Ltd, rejected a proposed pleading generated by artificial intelligence, labeling it 'AI slop' due to its poor quality.
• Justice Stewart dismissed the application for leave to file the further amended statement of claim, citing its repetitive nature, confusing structure, and non-compliance with Federal Court Rules 2011, including a lack of consecutively numbered paragraphs.
• The judge confirmed that while generative AI can be helpful, it must be used thoughtfully and critically to avoid producing 'slop' and causing prejudice or delay in proceedings.
• The court's decision aligns with its Use of Generative Artificial Intelligence Practice Note (GPN-AI) and underscores the need for rigorous human review of AI-generated legal documents.
• The underlying proceeding against the remaining respondents was not dismissed, indicating the applicant may still pursue a valid cause of action with a properly drafted pleading.

Justice Stewart stressed that genAI could produce 'slop' if litigants failed to use the tool thoughtfully and critically.

In a significant ruling highlighting the pitfalls of uncritical generative AI use in legal practice, Australia's Federal Court denied an application to file a further amended statement of claim, labeling the document as 'AI slop.' The decision, handed down by Justice Stewart on June 2, 2026, in the case of Barron v BT Funds Management Ltd (Leave to File) [2026] FCA 1220, underscores the judiciary's expectation for rigorous human oversight in court filings, particularly when leveraging artificial intelligence tools.

The applicant in the insurance matter, which involved claims related to salary continuance and total and permanent disability insurance, had initially filed an amended statement of claim against the Australian Financial Complaints Authority Ltd (AFCA), BT Funds Management Ltd (BTFM), and TAL Life Insurance Services (TAL). This initial document comprised 56 paragraphs across 16 pages. Subsequently, the applicant sought leave to file a *further* amended statement of claim, presenting a proposed pleading that was notably more extensive and problematic. This new draft spanned 33 pages, contained approximately 400 paragraphs, utilized a very small font, and was single-spaced. Crucially, it included 112 references to filenames stored on the applicant's personal computer, rather than properly cited evidence.

During the hearing, the applicant confirmed that generative artificial intelligence had been used to produce the proposed pleading. The two remaining respondents in the case, BTFM and TAL, opposed the filing of this document. Justice Stewart ultimately dismissed the interlocutory application for leave to file the proposed further amended statement of claim, citing its numerous deficiencies. However, the judge clarified that the underlying proceeding against the two remaining respondents would not be dismissed entirely. The court noted that the applicant might still be able to plead a reasonable cause of action against TAL, particularly concerning policy wording effective before January 1, 2018, which had been novated to TAL's predecessor, Westpac Life Insurance Services Ltd. If such a claim were viable, a corresponding claim against BTFM, based on its acceptance of TAL's reliance on incorrect policy wording, could also potentially proceed. No costs were awarded, as none were sought by the respondents.

Justice Stewart's ruling in Barron v BT Funds AI pleading denied offered a stern critique of the proposed document, describing it as 'highly repetitive, prolix, confusing, and very difficult to follow.' The judge explicitly referred to the pleading as 'AI slop,' a term he defined by quoting an AI platform: 'a disparaging term for low-quality content produced by generative AI, often rapidly and in large quantities, with little human judgement, checking or editing. It may be grammatically fluent or visually polished while being repetitive, vague, inaccurate, incoherent or devoid of substance.' This characterization directly implicated the lack of human judgment and critical review in the document's creation.

The Federal Court AI slop pleading was found to violate several procedural requirements. Specifically, the proposed pleading lacked consecutively numbered paragraphs in its initial pages, a clear breach of r 16(1)(a) of the Federal Court Rules 2011 (Cth). Furthermore, while some documents referenced in the further amended statement of claim were indeed part of the evidence, the judge observed that the proposed pleading's references were meaningless due to differing document names. These deficiencies, among others, led Justice Stewart to conclude that the document would likely cause prejudice, embarrassment, or delay in the proceeding, a ground for rejection under r 16.02(2)(d) of the Federal Court Rules 2011 (Cth). The court's decision also referenced its own Use of Generative Artificial Intelligence Practice Note (GPN-AI), indicating that the proposed pleading fell short of the expectations outlined in sections [1.3], [1.4], and [2.1] of that guidance.

The ruling by Justice Stewart on the generative AI legal drafting rejection serves as a critical reminder for legal professionals regarding the responsible integration of artificial intelligence into their workflows. While acknowledging that the proper and prudent use of genAI could potentially assist litigants in presenting their cases more effectively, Justice Stewart stressed that such tools are capable of producing 'slop' if not employed thoughtfully and critically. This highlights the imperative for lawyers and compliance officers to exercise extreme caution and rigorous human review when utilizing AI for court filings.

The Australia Federal Court GPN-AI, along with this specific ruling, reinforces the judiciary's stance that technological advancements do not diminish the fundamental requirements for clarity, accuracy, and adherence to procedural rules in legal documents. The case of Barron v BT Funds Management Ltd stands as a precedent, emphasizing that courts will reject poorly drafted, AI-generated documents that fail to meet these standards. This judicial perspective is not isolated; in a separate development, the Supreme Court of Victoria's Court of Appeal, on July 2, 2026, denied leave to appeal in an estate proceeding where the deceased's son had reportedly used AI to prepare a case that included references to non-existent or irrelevant authorities. Both decisions collectively underscore the judiciary's growing awareness of AI's capabilities and limitations, and the unwavering expectation for human accountability in legal work.

Lawyers and compliance officers must exercise extreme caution and rigorous human review when using generative AI for court filings, as courts will reject poorly drafted, AI-generated documents that fail to meet procedural rules and clarity standards. This ruling underscores the importance of thoughtful and critical application of AI tools in legal practice to avoid prejudice to a client's case and ensure compliance with court expectations.

Source: Original reporting via The Law Society of New South Wales