AI op-ed disclosure row hits Western Sydney University

Western Sydney University is reviewing how a senior academic integrity official used AI to help draft a newspaper opinion piece that warned students against using the same technology to cut corners.

The university confirmed pro vice-chancellor for quality and integrity Cath Ellis used Microsoft’s Copilot while preparing the article. It appeared in the Sydney Morning Herald and was later removed after questions about disclosure, The Guardian reported.

“To write her opinion article, Prof. Ellis uploaded 40,000 words of her own original materials into a Copilot Large Language Model (LLM).”
Western Sydney University spokesperson, via The Guardian

The case has turned AI use in education from a broad integrity debate into a practical governance problem for universities, publishers and senior staff. Whether staff can use generative tools in academic or public writing is only one part of it. The sharper test is what they must tell readers, who checks that disclosure, and whether senior staff face the same clarity demanded of students.

Ellis has argued that using a model to organise original material is different from outsourcing a piece to AI. She told The Age the article was “written with AI”, not by AI.

“It was written with AI, and there’s a really big difference there.”
Cath Ellis, via The Age

Disclosure becomes the test

That distinction may matter inside a university policy office. For readers, it is less tidy.

The article carried Ellis’s byline and warned students against taking shortcuts with AI, according to the reports. Once the drafting process became public, the standards question moved from technical authorship to trust.

Sydney Morning Herald editor Jordan Baker told The Guardian the situation was unacceptable and said the masthead was investigating further. The article’s removal shows how quickly AI disclosure can become a publishing issue, rather than an education-sector argument kept inside campuses.

Australian universities have spent the past three years rewriting assessment rules, plagiarism guidance and staff advice around generative AI. Much of that work has focused on students: what counts as assistance, what must be declared, and when chatbot help crosses into misconduct. The Ellis matter puts the staff side of that framework under the same light.

The dispute also lands while public confidence in AI companies remains weak. ABC News reported on Thursday that only 4 per cent of Australians surveyed trusted AI companies with their private information, while 1 per cent said they had complete trust that AI would be used responsibly.

Those figures do not decide what happened at Western Sydney University. They explain why disclosure now carries more weight. In a low-trust environment, institutional assurances depend less on saying AI was used responsibly and more on showing readers exactly how it was used.

For universities, the practical question is narrower and harder: whether staff guidance is explicit enough for public writing, media commentary and policy advocacy, as well as coursework and research. A rule that works for a student essay may not be enough for a senior official writing under a university title in a national newspaper.

Western Sydney University’s review will be watched for the standard it sets, not just the finding it reaches. If the institution treats the episode as a disclosure failure, other universities may need to tighten their own advice before the next AI-assisted byline becomes public.