The short version
- A Senate inquiry has highlighted significant citation errors in a report that justified Australia's social media ban for users under sixteen.
- The organization behind the study admitted to using artificial intelligence for editing after investigators found metadata linking sources to ChatGPT.
- Government officials have not independently verified the accuracy of the citations, citing the difficulty of retroactively checking thousands of references.
A Senate inquiry into strengthening Australia’s social media restrictions has uncovered serious flaws in a pivotal report that helped justify the implementation of an age ban for users under sixteen. The investigation revealed that sections of the document contained citations to academic articles that do not exist, while others linked to incorrect papers or misattributed authorship. These discrepancies have sparked concerns about the reliability of the technical evidence used to shape federal policy on digital safety.
The report in question was produced by the Age Check Certification Scheme, a UK-based organization contracted for a $3.48 million trial last year. The study evaluated various age-verification technologies intended to help social media platforms comply with the new law, which took effect in December of the previous year. At the time, Communications Minister Anika Wells praised the findings, stating they demonstrated effective options for age assurance and paved the way for the ban’s enforcement.
While the majority of the report relied on direct testing of available verification tools, a specific chapter covering emerging technologies cited numerous journal papers to support potential new methods. A submission to the ongoing Senate inquiry alleged that these citations appeared to be hallucinations generated by artificial intelligence rather than references to genuine sources. The inquiry sought to determine whether the foundational research supporting the ban’s technical feasibility was sound.
Initial responses from the Age Check Certification Scheme denied any use of AI in generating the report or its cited materials. A spokesperson insisted that every link had been manually checked for relevance and authenticity. However, after investigators identified metadata within several links indicating they originated from ChatGPT, the organization conceded that AI tools were used to rewrite paragraphs for conciseness. They maintained that this editing did not compromise the integrity of the research or the verification process.
Independent analysis by news outlets identified at least six distinct reference errors in the section on emerging technologies. These included Digital Object Identifiers that led to non-existent papers, links pointing to unrelated articles, and combinations of author names and publication years that matched no known records. In one instance, a cited paper was claimed to have been accessed in March 2025, yet the lead author confirmed the study was not publicly available until June of that year. The author also noted that the person listed as the lead researcher by the certification scheme had never worked on the paper.
Further scrutiny revealed inconsistencies in the organization’s attempts to clarify these errors. When asked to provide correct references for a citation attributed to Weber et al. from 2011, the scheme provided a Monash University study. However, investigators found no reference to Weber within that document. Additionally, a list of corrected journal articles included years and titles that did not align with the original citations, suggesting a persistent lack of accuracy in the documentation.
The controversy echoes previous incidents where government contractors faced backlash for using AI without proper disclosure. Last year, Deloitte refunded part of a $440,000 contract after admitting to AI use that resulted in errors. The federal government has expressed skepticism toward such practices, emphasizing the need for rigorous human verification in official reports. Questions directed to Minister Wells’ office were redirected to her department, which stated it was examining the concerns and would engage with the certification scheme as appropriate.
During a recent hearing, department officials noted that they had met with representatives from the Age Check Certification Scheme, who attributed the errors to broken links that had previously functioned correctly. Officials stated they had not independently verified these claims, describing the task of retroactively checking thousands of citations in a 1,000-page report as extremely difficult. They characterized the issues as a small number of errors within a large volume of data, though the implications for policy credibility remain under review.
The Senate inquiry continues to assess the strength of the legal and technical framework surrounding the social media ban. The discovery of fabricated or erroneous citations raises broader questions about the oversight mechanisms for government-commissioned research. As lawmakers consider further strengthening the legislation, the integrity of the underlying evidence has become a central point of contention.
Stakeholders are now calling for greater transparency in how AI tools are utilized in public sector reporting. The incident underscores the challenges of integrating advanced technology into policy development without compromising accuracy. Until the discrepancies are fully resolved and verified, the technical basis for Australia’s age-restriction measures remains subject to significant doubt.
Sources behind this briefing
Go to the original reporting
- The Guardian World↗Report supporting Australia’s teen social media ban appears to contain AI hallucinations, Senate hears
