AI-Generated Fake Research Is Flooding Australia's Parliamentary Inquiry System
Submissions to government committees are citing studies that don't exist and attributing them to real academics, with some reports built largely on invented sources.
Key points
- Guardian Australia found dozens of parliamentary submissions containing AI-invented sources attributed to real researchers.
- The problem spans the political spectrum, from progressive to conservative advocacy groups.
- Some parliamentary committee reports have cited submissions where most references appear to be fabricated.
- The issue stems from AI "hallucinations", where a large language model confidently produces false information that looks real.
- No Australian law currently bars AI-generated content from the parliamentary submission process.
Australia's parliamentary inquiry system works like a formal suggestion box for democracy. Citizens, researchers, businesses and lobby groups submit written evidence to government committees, which lawmakers then use to shape laws and policy. It is supposed to be grounded in fact.
According to an investigation by Guardian Australia, it increasingly is not.
The outlet found dozens of submissions, spread across the political spectrum, that cite research papers which do not exist, attribute quotes to academics who never said them, and point to journals that have no record of the claimed study. In several cases, committee reports then cited those submissions as credible evidence, passing the invented material one step closer to actual law.
What exactly is a hallucination?
A hallucination is what researchers call it when an AI confidently produces something false. The AI, typically a large language model (a type of software trained on vast amounts of text, and the technology behind tools like ChatGPT), does not know it is wrong. It generates plausible-sounding citations the same way it generates plausible-sounding sentences: by predicting what text should look like, not by checking facts.
The result can look authoritative. A footnote might name a real university professor, a real journal, and a plausible study title. Only when someone searches for the paper does the fiction unravel.
For busy committee staff reading hundreds of submissions, that search does not always happen.
Why does this matter to ordinary Australians?
Policy built on invented evidence is a real problem. If a committee recommends tightening drug regulations, loosening environmental rules, or changing school curricula partly because of research that never existed, the downstream effects land on patients, communities and children.
It also corrodes trust. Academics whose names are attached to studies they never wrote have no way of knowing it is happening unless someone checks. Real researchers who submitted genuine, carefully sourced evidence now compete for attention against fabricated content that looks just as polished.
| Submission problem | What it means |
|---|---|
| Invented study title | Paper does not exist anywhere |
| Wrong author attribution | Real person, fake quote or finding |
| False journal citation | Journal exists but never published the claimed article |
| Hallucinated statistics | Numbers presented as research findings with no source |
Guardian Australia's analysis found the issue is not confined to one side of politics. Submissions from advocacy groups across the ideological spectrum contained fabricated material, suggesting this is a tool of convenience rather than coordinated bad faith.
What happens next?
Australia's parliamentary committees have no current rule requiring submitters to declare whether they used AI, and no systematic process for verifying citations. Some committee chairs are now reportedly discussing disclosure requirements, though nothing is in place yet.
For now, the safest assumption for anyone reading a policy submission is the same one that applies to anything written with AI assistance: check the primary source before trusting the claim. If the paper cannot be found, it may not exist.
Common questions
Can AI-generated submissions be banned?
No blanket ban exists in Australia yet. Committees are exploring disclosure rules, where submitters would have to flag AI involvement, but enforcement would be difficult.
How can I tell if a submission used AI to invent sources?
Search for the cited paper title plus the author's name in Google Scholar or a university library database. If it does not appear, treat the citation as unverified and look for the committee's own source-checking notes.
