An exclusive Guardian Australia analysis has revealed that Australia’s parliamentary inquiry process is being inundated with AI-generated material that invents studies and attributes non‑existent research to real academics and authors. The findings raise serious concerns about the integrity of evidence‑based policymaking, with some committee reports now citing submissions in which the majority of sources appear to be AI‑hallucinated.
Guardian Australia’s investigation found that dozens of submissions to federal inquiries – from across the political spectrum – incorrectly summarise real research or cite sources that do not exist. Large language models (LLMs) can generate text that appears credible but is entirely fabricated, a phenomenon known as ‘hallucination’.
The problem has infiltrated the very system designed to help parliament make better decisions by hearing from experts and constituents. In some instances, committee reports have relied on submissions where most sources are likely AI-generated hallucinations, meaning that false information could directly shape parliamentary recommendations and legislation.
The Guardian did not name specific submissions or committees, but the analysis underscores a systemic vulnerability in Australia’s democratic process. As AI tools become more accessible, the risk of unintentional – or deliberate – introduction of fake data into official records grows.
The findings come amid broader global concern about generative AI’s impact on information integrity. In Australia, there is currently no formal mechanism to screen parliamentary submissions for AI-generated content or to verify the accuracy of cited sources.
No individuals or groups were quoted in the Guardian’s report, and none of the affected submitters have been publicly identified.
Analysis
Why This Matters
- Parliament’s inquiry process relies on accurate submissions to inform policy. AI‑hallucinated sources could lead to flawed legislation based on non‑existent evidence.
- The credibility of Australia’s democratic institutions is at stake if the public perceives that unverified AI content is shaping government decisions.
- This trend could waste significant taxpayer resources, as parliamentarians and committee staff must now treat all submissions with heightened scepticism and invest time in source‑checking.
Background
Generative AI tools such as large language models are capable of producing coherent‑sounding text that is not grounded in fact. A common failure mode is ‘hallucination’, where the model invents plausible‑looking references, dates and study titles. As these tools become widely available, governments worldwide are grappling with their potential to distort public discourse and administrative processes. Australia’s parliamentary submission system, which allows individuals and organisations to contribute to inquiries, has traditionally assumed good faith and human authorship – an assumption now under threat.
Key Perspectives
Parliamentary committees: They rely on submissions for expert and public input. Without a vetting system, committees risk citing fabricated research, undermining the quality of their reports and recommendations.
Submitters (legitimate experts and citizens): Honest participants may find their contributions devalued as scrutiny increases. Some submitters may unknowingly use AI tools that hallucinate sources, believing they are submitting accurate work.
Critics and transparency advocates: They argue that the discovery exposes a critical gap in Australia’s democratic safeguards. They will likely call for mandatory AI‑disclosure labels on submissions and formal verification protocols.
What to Watch
- Whether parliamentary committees introduce explicit rules requiring submitters to declare use of AI tools.
- Potential government or cross‑party investigations into the scale of AI‑generated content across all federal inquiry processes.
- Broader legislative moves to regulate the use of generative AI in official public consultations, both in Australia and internationally.