Australia's parliamentary inquiry system is being flooded with AI-generated submissions containing invented studies and fake references attributed to real academics. Guardian Australia identified at least 39 submissions to current parliament inquiries that contain what appear to be hallucinated citations, raising the risk that lawmakers will make policy decisions based on evidence that does not exist.
The problem stems from large language models (LLMs) such as ChatGPT and Claude, which can produce plausible but entirely fabricated references. These AI hallucinations occur because the systems predict the next most likely text based on training data rather than verifying accuracy. The issue is compounding: Google's AI summaries will sometimes create summaries of fake references as though they were genuine, and both Google and ChatGPT can cite inquiry submissions containing those fake references, creating a cycle of misinformation.
Real academics, nonexistent research
One submission to an inquiry into family violence and suicide included a hallucinated reference attributed to Divna Haslam, a University of Queensland associate professor and clinical psychologist. The document misstated her team's research findings. Haslam said the reference looked so real that someone taking a cursory glance could be convinced it was legitimate.
"It's very frustrating to have invested time, money, effort, expertise in rigorous research," she said. "Then to see something that's inaccurate and inappropriately attributed anyway is really concerning."
The submission's author, Drilldown Reports, acknowledged using AI in its research process and said it had identified the errors in a follow-up submission, but the correction deadline had passed. A spokesperson attributed the mistake to human error in uploading the wrong document, not the AI itself.
Nicole Gurran, a professor of urban and regional planning at the University of Sydney, also found her name attached to nonexistent work in a housing inequity inquiry submission. "Fake citations, even if an accidental 'collage' where the claim is correct but the chain of reference to evidence is broken, undermines that," she said, referring to the transparency and contestability that citations provide.
How the fake submissions were identified
Guardian Australia built a program that extracted references from all inquiry submissions made to the current parliament and checked them against CrossRef and Google Scholar. Documents where 20% or more of references could not be matched were manually verified. Links were also checked for ChatGPT metadata tags, which persist when copied into documents.
The findings represent a conservative estimate. The method only identifies submissions with incorrect citations, meaning documents that used AI to generate text without references would go undetected. More than 100 papers included ChatGPT URL tags in reference links, indicating how widely LLMs are being used.
Christian Downie, a professor in the Australian National University's school of regulation and global governance, said the implications extend beyond individual bad decisions. "We're also starting to undermine the public's trust and confidence in the types of institutions that underpin our democracy," he said.
Guidance and responsibility
Senate advice for inquiry submissions warns that AI use can present risks to information quality and that accuracy remains the responsibility of the submitter. Downie said the inquiry process needs to remain open, but new guidelines may be needed to encourage truthfulness.
OpenAI said users should treat ChatGPT as "a first draft, not a final source" and verify quotes, data, and references. A Google spokesperson said AI Overviews operate like traditional search by matching web content to query terms. Neither company addressed the specific issue of their systems citing inquiry submissions that contain fabricated references.
For government professionals working with inquiry processes, understanding how Generative AI and LLM systems produce hallucinations is becoming a core competency. The tools are now common enough in submissions that verification skills are no longer optional.
Why this matters for government professionals
Committee staff, policy officers, and parliamentarians now face a practical challenge: distinguishing legitimate expert evidence from AI-generated text that looks credible but cites research that does not exist. The submissions identified by Guardian Australia came from individuals and organisations across the political spectrum, and many authors were unaware AI could fabricate references. For anyone who reviews, summarises, or acts on inquiry submissions, checking citations against academic databases before relying on them is no longer a formality - it is a necessary safeguard against making decisions on evidence that was never real. Building that verification step into standard workflow, and investing in AI for Government training, is the most direct way to protect the integrity of the inquiry process.
Your membership also unlocks:
