Report's AI References Questioned After Errors Found

A comprehensive report assessing age verification technologies for Australia's social media ban is facing questions regarding its integrity. The report, funded by a $3.48 million trial, utilized various technologies to test potential age checks for individuals under 16, influencing the ban's December implementation. Concerns arose when a Senate inquiry submission highlighted citation errors, suggesting AI hallucinations.

Initially, the report's authors denied using AI for generating content or citations. However, they later conceded that AI, specifically ChatGPT, was employed to rewrite paragraphs for conciseness. Despite this admission, they maintained that human verification ensured all links and sources were genuine.

Further analysis revealed at least six erroneous references, including non-existent Digital Object Identifiers (DOIs) and incorrect paper associations. One cited academic paper was confirmed to be inaccessible before its stated publication date, with its listed lead author unassociated with the work.

Explanations provided by the authors to rectify these errors also contained inaccuracies and outdated information. The department overseeing the trial stated it is examining the concerns and engaging with the authors. Experts warn that such inaccuracies in government reports could lead to flawed decisions and erode public trust.