Report backing australia's teen social media ban used ai, senate told

A report commissioned to test age-verification technology for Australia's social media ban for teenagers has come under scrutiny after a Guardian Australia investigation revealed multiple citation errors, including links to non-existent academic papers. The report, produced by the Australian Communications and Media Authority's (ACMA) contractor, the Australian Centre for Cyber Security (ACCS), was cited by Communications Minister Anika Wells as demonstrating effective options for age-checking, paving the way for the ban's implementation in December 2025. The errors, discovered in a 1,000-page report, include Digital Object Identifiers (DOIs) that link to papers that do not exist, DOIs pointing to incorrect papers, and author name, journal, and publication year combinations that do not match any known references. In one instance, a DOI linked to a paper that did not say what the report suggested it said. The Guardian also identified four links in parts E and K of the report containing metadata identifying ChatGPT as the source of the link.

When confronted with the findings, ACCS initially denied using AI in the generation of the report or cited materials, claiming each reference was checked as genuine and relevant. However, after the Guardian presented evidence of ChatGPT metadata in the links, the organization conceded that AI had been used for editing, specifically to rewrite some paragraphs more succinctly. They maintained that the inclusion of ChatGPT metadata in the links constituted disclosure of its use and, since the links were correct, they did not see a problem. The spokesperson later provided clarifications and new references, but this response also contained errors. One lead author of a cited study confirmed their article was not publicly accessible until June 2025, that the person originally cited as lead author had never been on the paper, and that the ACCS report had summarized their paper incorrectly.

The matter was raised during a Senate inquiry hearing, where department officials stated they met with ACCS, who claimed the errors were due to links breaking that had previously worked. The officials admitted they had not independently verified this claim, stating it would be very difficult to go back and check, and characterized the issue as a handful of errors in 26 pages of citations. The federal government has previously taken a dim view of contractors using AI in reports. A spokesperson for the department said they were examining the concerns and would engage with ACCS as appropriate. Prof Christian Downie from the Australian National University warned that false citations in government submissions, whether AI-generated or not, could lead to bad decisions and erode public confidence and trust in institutions, suggesting contractual arrangements and penalties may be needed to discourage such behavior.