The AU$3.48 million report was "polished" by ChatGPT: The technical basis for Australia's social media ban was backfired by false quotes. Australia spent AU$3.48 million to commission an age verification technology test, but was embarrassed by a report that was adulterated with AI.

The AU$3.48 million report was "polished" by ChatGPT: The technical basis for Australia's social media ban was backfired by false references. Australia spent AU$3.48 million to commission an age verification technology test, but was embarrassed by a report that was adulterated with AI. ACCS, the British agency responsible for the test, initially denied using artificial intelligence until the Guardian discovered the ChatGPT metadata in a link to the report, admitting that it had rewritten some passages with AI. Fake citations surface The core content of the key report that paved the way for Australia's social media ban for under-16s comes from real technology testing, but the literature citations in the "Emerging Technologies" chapter are full of holes. The Guardian checked and found that six references in the same chapter alone were incorrect: some DOIs pointed to non-existent papers, some linked to wrong documents, and the combination of author, journal, and year did not match any known research. What's even more outrageous is that the "error correction list" subsequently submitted by ACCS also has discrepancies in years and authors. The main author of one of the papers clearly stated that he had never participated in the research, and the paper was only published after the access time he claimed. Government trust has been hit hard again. ACCS insists that AI is only used to make expressions more concise and the links themselves correct, so it is not a problem. However, government contractors have a long history of using AI to mix water. Last year, Deloitte refunded A$440,000 in contract payments due to errors in reporting AI hallucinations. The Australian federal government has a negative attitude towards such actions. Academics have warned that false citations, whether caused by AI or not, may mislead government decision-making and erode public trust, and some lawmakers have demanded an apology from the contractor and refunds. A mere 26 pages of citations in a thousand-page report, but a defense that reveals "a few errors", may not be able to quell the crisis of trust surrounding AI ghostwriting. via AI News (author: AI Base)