The report that helped pave the way for Australia's under-16 social media ban appears to contain citations fabricated by artificial intelligence, according to a Guardian analysis published on Monday — an allegation that has now reached a Senate inquiry.
The A$3.48 million age-assurance technology trial was run last year by the UK-based Age Check Certification Scheme (ACCS) and tested technologies social media platforms could use to verify ages under the ban. Communications Minister Anika Wells heralded the report as evidence of "many effective options", and the ban took effect in December.
But in a submission to a Senate inquiry seeking to strengthen the law, concerns were raised that citations in the report's "emerging technologies" chapter appeared to be "AI hallucinated, rather than being based on real sources". Guardian Australia's own analysis identified at least six problematic references in that section: DOIs linking to papers that do not exist, DOIs pointing to the wrong papers, author–journal–year combinations matching no known publication, and links to papers that said nothing of what the report claimed.
When contacted, ACCS initially denied using AI, then conceded that ChatGPT had been used to rewrite paragraphs more succinctly — after the Guardian found metadata in four links identifying ChatGPT as their source. The organization said all links were "checked and verified" by humans, and that errors were inadvertent. A follow-up list of corrected references itself contained errors: one article was cited as accessed in March 2025 even though a lead author confirmed it was only published in June 2025, and the person listed as its lead author had never been on the paper.
At a Senate hearing on Friday, department officials said they had met with ACCS, which claimed the errors were due to links that had previously worked breaking over time — an explanation officials have not independently verified. They described the problems as a "handful of errors" in 26 pages of citations within a roughly 1,000-page report.
The episode echoes last year's Deloitte scandal, in which the consultancy refunded part of a A$440,000 government contract after errors were traced to AI use. Independent Senator Fatima Payman said the government should demand an apology and refund from ACCS, warning that unchecked AI citations erode public trust in institutions.
Experts say the case is a warning sign for governments worldwide as contractors increasingly use generative AI in official reporting — and a reminder that a polished citation list is not proof of a real source.
