Report supporting Australia’s teen social media ban appears to contain AI hallucinations, Senate hears
TL;DR
A Guardian analysis found that a report testing the technology behind Australia's under-16 social media ban contains links to academic articles that do not exist. The authors have conceded that ChatGPT was used during editing, but they deny that the citation errors were caused by AI hallucinations. The $3.48m age assurance technology trial was run by the UK-based Age Check Certification Scheme and tested various methods social media platforms could use for age verification. The issue was raised in a Senate hearing.
Nauti's Take
Using AI in the editing loop is legitimate and a real opportunity for speed, as long as every citation is checked by a human afterwards. The risk is on display here: invented references inside a document that legislation rests on turn efficiency into liability.
Anyone producing studies or policy papers needs a documented source check as a fixed step before publishing.
Sources