Don't outsource your thinking, however tempting that may be.
A professor who built her reputation on academic integrity published a column urging students not to outsource their thinking to AI — only for the column itself to have been drafted with AI assistance. When the Sydney Morning Herald discovered the undisclosed use and removed the piece, it illuminated a contradiction that extends far beyond one academic: as generative AI quietly enters the workflows of writers, educators, and institutions, the line between using a tool and becoming dependent on it grows harder to draw, and harder still to police.
- A senior academic charged with upholding university integrity published a warning against AI shortcuts while having used AI to write the very warning — a contradiction that collapsed the argument the moment it was discovered.
- The Sydney Morning Herald pulled the column from its website after its editor called the undisclosed AI use 'unacceptable,' triggering a formal investigation into a breach of the publication's editorial standards.
- Western Sydney University pushed back, framing the AI use as sophisticated and legitimate — arguing that feeding a decade's worth of original research into a language model is meaningfully different from simply asking it to write something from scratch.
- The incident joins a growing list of disclosure failures across Australian and international media, suggesting that undisclosed AI use is not an isolated lapse but a systemic pressure point as newsrooms scramble to define what transparency actually requires.
- The unresolved question — where does AI assistance end and AI authorship begin — now hangs over every byline, with no consensus in sight and the tools themselves offering no easy answer.
Prof. Cath Ellis, pro vice-chancellor for quality and integrity at Western Sydney University, published an opinion piece in the Sydney Morning Herald urging students not to outsource their thinking to AI. The column was a direct response to fellow academic Kylie Moore-Gilbert's concern that students were increasingly being graded on their ability to prompt AI rather than on genuine learning. Ellis acknowledged the problem but argued that authentic effort still mattered and that university remained worthwhile.
The piece unraveled when an AI-detection service flagged it as machine-generated. SMH editor Jordan Baker, upon learning that Ellis had used a Copilot language model to draft the column without disclosure, removed it from the website and called the omission unacceptable.
Western Sydney University defended her, explaining that Ellis had fed 40,000 words of her own accumulated research into the model, which then summarized and generated draft material from that knowledge base. The university called this a sophisticated and appropriate use of generative AI, and noted that detection tools can identify machine involvement but cannot assess whether that involvement was justified.
The Herald's parent company, Nine, permits AI for research and idea generation but prohibits it from writing publishable stories, and requires clear labeling when AI-generated material appears in print. Neither Ellis nor the university had informed the Herald of the AI's role.
The episode is not isolated. In March, Crikey removed multiple articles after discovering undisclosed AI use, and the New York Times cut ties with a freelancer who admitted to AI assistance on a book review. Ellis's column — intended as a caution against intellectual shortcuts — became instead an emblem of the unresolved tension between using AI as a tool and allowing it to speak in one's voice.
Prof. Cath Ellis, a senior academic at Western Sydney University, published an opinion piece in the Sydney Morning Herald last month that carried an unmistakable message: students should resist the temptation to outsource their thinking to artificial intelligence. "Don't cut corners. Don't outsource your thinking, however tempting that may be," she wrote, responding to concerns raised by fellow academic Kylie Moore-Gilbert about students gaming their education through AI-generated work. The irony emerged when the piece itself was run through an AI-detection service and flagged as machine-generated.
Ellis, who holds the title of pro vice-chancellor for quality and integrity at the university, had written her column as a counterpoint to Moore-Gilbert's warning that students were increasingly "being graded on who can write the best AI prompts" rather than on genuine learning. Ellis acknowledged the problem was real but argued that university remained worthwhile and that authentic effort would ultimately distinguish itself. When the Sydney Morning Herald's editor Jordan Baker learned the column had not been disclosed as AI-assisted work, he removed it from the website, calling the omission "unacceptable."
Western Sydney University defended Ellis's approach in a statement to Guardian Australia. The university said Ellis had fed 40,000 words of her own original materials—accumulated over more than a decade as a global expert in her field—into a Copilot language model, which then summarized and generated prompts based on that knowledge base. The university characterized this as "a sophisticated and appropriate use" of generative AI, arguing that the technology had simply distilled her existing expertise into draft form. A university spokesperson noted that AI-detection tools like Pangram can identify when a machine has been involved in writing but cannot judge whether that involvement was justified or misused.
The Sydney Morning Herald's parent company, Nine, maintains a policy that permits writers to use AI for research and idea generation but explicitly prohibits AI from being used to write stories for publication. The policy does allow for unlabeled use of "assistive AI" in certain contexts, but requires clear labeling when AI-generated material appears in print. Neither Ellis nor the university informed the Herald that generative AI had been used to draft the column. Baker said in a subsequent statement that the Herald was investigating the breach and reaffirmed that the piece did not meet its editorial standards.
The incident sits within a broader pattern of disclosure failures across media organizations as generative AI tools proliferate in newsrooms. In March, the Australian publication Crikey removed multiple articles from a writer after discovering undisclosed AI use in proofreading. That same month, a freelance journalist admitted to using AI assistance on a book review that bore similarities to a Guardian piece, prompting the New York Times to sever its relationship with him. As AI becomes more embedded in editorial workflows, the question of what constitutes appropriate use—and what must be disclosed—remains unsettled. Ellis's column, meant to warn students against intellectual shortcuts, instead became a case study in the murky space between using AI as a tool and using it to generate publishable work.
Notable Quotes
The Herald was not informed of the use of AI in the compilation of the article by either the author or Western Sydney University. Clearly this is unacceptable.— Jordan Baker, editor of the Sydney Morning Herald
The University believes the AI use in this case was appropriate, as it involved uploading the author's own 40,000 words of original materials to summarize her expertise.— Western Sydney University spokesperson
The Hearth Conversation Another angle on the story
Why does it matter that an academic used AI to write about not using AI? Isn't that just ironic?
It's more than irony. It's about institutional credibility. Ellis holds a position overseeing quality and integrity. When she tells students not to cut corners, that message carries weight only if she's visibly not cutting corners herself.
But the university says she fed her own 40,000 words into the system. Isn't that just using a tool to organize her thinking?
That's the defense, yes. But there's a difference between using AI to organize your notes and submitting AI-drafted prose to a newspaper without saying so. The Herald has a policy. The university knew the policy. They didn't disclose.
So this is really about transparency, not about whether the AI use itself was wrong?
Exactly. The university might be right that feeding your own expertise into an LLM is legitimate. But that legitimacy evaporates the moment you publish the output without telling readers what happened.
Why does the Herald care? Readers got her actual ideas either way.
Because readers deserve to know how the words reached them. If you're reading an argument about AI in education, you want to know whether the author wrote it herself or had a machine draft it. That's not a small thing.
What happens now?
The Herald is investigating. Other newsrooms are watching. The real question is whether media outlets will tighten their policies or whether this becomes normalized—another disclosure failure in a long list.