Updated
Updated · The Guardian · Aug 17
Australian Senate Probes $3.48 Million Social Media Ban Report Over AI Citation Errors
Updated
Updated · The Guardian · Aug 17

Australian Senate Probes $3.48 Million Social Media Ban Report Over AI Citation Errors

2 articles · Updated · The Guardian · Aug 17

Summary

  • Six references in the age-assurance report backing Australia’s under-16 social media ban contained errors, and senators heard allegations that some citations may have been AI-hallucinated.
  • ACCS, the UK-based contractor behind the $3.48 million trial, conceded ChatGPT was used to rewrite paragraphs and that links carried ChatGPT metadata, but denied AI generated the research or citations.
  • Guardian Australia found broken or mismatched DOIs, incorrect author and journal details, and a paper cited as accessed in March 2025 even though its lead author said it was not public until June 2025.
  • Department officials told a Senate hearing ACCS said the links worked when published and later broke, but the department has not independently verified that claim and is examining the concerns.
  • The dispute revives scrutiny of AI use by government contractors after Deloitte refunded part of a $440,000 contract last year, with critics warning false citations could distort policy decisions and erode trust.

Insights

If the experts certifying online age checks used fake AI citations, can any government tech policy be trusted?
Did a multi-million dollar AI hallucination just break Australia's highly touted teenage social media ban?