Democracy at Risk: Australian Government Inquiries Plagued by AI-Generated Fabrications
Australia’s parliamentary inquiry system—a cornerstone of the nation’s democratic process designed to gather expert evidence and inform critical legislation—is facing an unprecedented integrity crisis. A new investigation reveals that the system is being flooded with AI-generated submissions containing “hallucinated” studies, nonexistent research, and falsely attributed academic credentials.
These AI models, which operate by predicting the next most likely string of text rather than verifying factual accuracy, are creating documents that appear authoritative but are fundamentally deceptive. Experts warn that if left unchecked, these fabrications risk subverting the policy-making process by forcing parliamentarians to base laws on evidence that quite simply does not exist.
The Growing Cycle of Misinformation
The threat has been compounded by modern search engine integration. Where a researcher might have previously realized an academic reference was fake after a Google search failed to return results, current AI-powered search summaries now frequently “confirm” these hallucinations. By summarizing fake papers as if they were genuine, AI tools are creating a self-reinforcing loop of misinformation that is increasingly difficult for committee staff to police.
Christian Downie, a professor at the Australian National University’s School of Regulation and Global Governance, warns that the consequences extend far beyond simple technical errors. “We’re starting to undermine the public’s trust and confidence in the types of institutions that underpin our democracy,” Downie noted.
Academics Targeted by Ghost References
The impact is personal for many researchers. Associate Professor Divna Haslam, a clinical psychologist at the University of Queensland, recently discovered her own name attached to a hallucinated reference in an inquiry submission regarding family violence and suicide.
“It’s very frustrating to have invested time, money, effort, and expertise in rigorous research, then to see something that’s inaccurate and inappropriately attributed,” Haslam said. She cautioned that the risks of such misinformation in sensitive policy areas, such as domestic violence, are dangerously high.
In another instance, Margaret Simons, a prominent journalist and academic, found herself “cited” in a submission to a Senate inquiry on climate change. The document claimed she had authored a specific paper critiquing The Guardian—a piece that does not exist. Even for the subject themselves, the fabrication was so convincing that it prompted a moment of self-doubt.
A System Under Pressure
An investigation into recent parliamentary submissions identified at least 39 documents containing likely AI-generated hallucinations. Because this methodology only flags broken citations, experts believe the true scale of AI-generated content in the parliamentary record is likely much higher.
While some submitters have claimed these issues were the result of “human error” during the submission process, the persistence of the trend suggests a lack of rigorous fact-checking. This follows a high-profile incident in 2025 where the consultancy firm Deloitte was forced to issue a partial refund to the Albanese government after a $440,000 report was found to contain fabricated court cases and sources.
As the AI-generated material continues to permeate public discourse, the burden of verification has shifted squarely onto the shoulders of parliamentary committees.
OpenAI, the creators of ChatGPT, maintains that users should treat their output as “a first draft, not a final source,” placing the responsibility for verifying references back on the user. Meanwhile, with democracy relying on the quality of evidence submitted for public record, calls are growing for new, more stringent guidelines to ensure that the voices shaping Australia’s future are not being written by machines.
