A Guardian Australia investigation has found that Australia's parliamentary inquiry system is being flooded with AI-generated submissions containing hallucinated content, including invented studies and fabricated research wrongly attributed to real academics. Some committee reports have cited submissions in which most sources appear to be AI "hallucinations" — plausible but false content generated by large language models such as ChatGPT or Claude. Experts warn this risks parliamentarians making policy decisions based on evidence that does not actually exist, undermining the integrity of a process designed to inform lawmaking with genuine expert and public input.

The problem is compounded by AI tools like Google's search summaries, which can generate convincing-sounding descriptions of fake references rather than flagging them as non-existent, creating a feedback loop of misinformation. One submission to an inquiry into family violence and suicide falsely attributed a hallucinated reference and misrepresented findings from University of Queensland researcher Divna Haslam, who described the fabrication as "scary" for how authentic it appeared. The submission's author, Drilldown Reports, said it used AI in its research process, had identified the errors too late to correct them before the deadline, and attributed the mistake to human error in uploading rather than AI itself. The committee chair overseeing the inquiry said all submissions are scrutinised as part of the standard evidence-gathering process.