Incident Overview
On May 31, 2026, Nine Publishing’s Sydney Morning Herald and The Age retracted an opinion article after discovering it had been generated with Microsoft Copilot without proper disclosure. The article, authored by Western Sydney University academic Cath Ellis, was withdrawn following an internal review that found it violated the publications’ AI usage policies. Ellis admitted to using Copilot to structure her own notes, though she did not disclose this to the editorial team. The mastheads emphasized that the article failed to meet their editorial standards regarding transparency in AI-assisted writing.
The retracted article, titled ‘Odd choices of words’: How an academic’s AI use was exposed by her peers, was published before the breach was identified. Its removal followed an investigation triggered by suspicions raised by peers regarding unusual phrasing and stylistic inconsistencies in the text.
The retracted version of the article is archived and accessible as an image through the AI Incident Database (AIID). This incident is categorized as a variant of broader AI policy violations in newsrooms, where undisclosed AI assistance compromises journalistic standards and public trust.
Possible Root Cause
The primary root cause of this incident appears to stem from a lack of clear communication between the author and the editorial team regarding AI tool usage. Cath Ellis, a professor specializing in academic integrity, stated she used Microsoft Copilot to organize and refine her notes before drafting the article. However, she did not inform the Sydney Morning Herald or The Age of this assistance, which contradicted the publications’ stated policies on AI disclosure.
The editorial teams of both mastheads later clarified that the article did not comply with their internal guidelines, which mandate explicit disclosure of AI-assisted writing.
While many news organizations have adopted policies to regulate AI use, their effectiveness depends on consistent adherence by contributors, particularly freelancers, academics, or external writers who may not be fully versed in the publications’ editorial protocols. In this case, the academic’s expertise in integrity-related fields may have inadvertently contributed to the oversight, as she likely assumed her use of Copilot for structural assistance was innocuous.
Takeaways
- Editorial Policy Enforcement Gaps: The incident reveals that even well-established publications with explicit AI disclosure policies may struggle with enforcement, particularly when working with external contributors.
- Reputational and Epistemic Risks: Undisclosed AI assistance in journalism poses dual risks: reputational damage to the publication and erosion of epistemic integrity. Readers and peers may question the authenticity of the content, as seen in this case where stylistic inconsistencies raised suspicions. For opinion pieces, where the author’s perspective is critical, transparency in AI use is non-negotiable to maintain credibility.
- Author Accountability and Training: Contributors, especially those from academic or research backgrounds, may underestimate the implications of AI tool use in professional writing. News organizations should provide clear guidelines and training on AI disclosure, emphasizing that even structural or organizational assistance must be declared. This incident serves as a cautionary example for authors who assume minor AI use is exempt from transparency requirements.
- Broader Industry Trends: This case is part of a larger pattern of AI policy breaches in 2026, reflecting broader challenges in integrating generative AI into journalism while preserving ethical standards.
- Technical and Ethical Audits: Publications may benefit from periodic audits of AI tool usage, particularly in high-profile or opinion-driven content.
The incident serves as a case study in the unintended consequences of AI tool adoption in professional writing, where good intentions—such as improving efficiency—can clash with ethical obligations.
Sources
Incident summary from the AI Incident Database, used under CC BY-SA 4.0.