TL;DR
Multiple governments have faced public embarrassment due to AI-generated false citations and errors in official documents. These incidents reveal the risks of using AI without proper verification, raising concerns about accountability and trust.
Five instances over the past two years have demonstrated how AI hallucinations have led to embarrassing errors in government documents, raising questions about the reliability of AI in official settings. These incidents involve errors in citations, fabricated references, and false data, highlighting the risks of deploying AI without adequate human oversight.
The first confirmed case involved South Africa’s Draft National Artificial Intelligence Policy, published in April 2025, which contained at least six AI-generated fake sources in its bibliography. Minister Solly Malatsi acknowledged the inclusion of AI hallucinations, leading to the policy’s withdrawal 17 days after publication. This marked the first time a government withdrew a document due to AI-generated errors.
In the United States, a May 2025 report on children’s health titled “Make America Healthy Again” included numerous incorrect citations, some referencing non-existent studies. White House spokesperson Karoline Leavitt dismissed these as “formatting issues,” but a corrected version was uploaded shortly after.
Australia’s Department of Employment and Workplace Relations commissioned a report from Deloitte in August 2025, which was later found to contain fabricated references and quotes due to AI use. Deloitte acknowledged the errors, corrected the report, and refunded the government $290,000 of the $440,000 paid for the report.
Similarly, in Canada, Deloitte produced a 526-page healthcare report for Newfoundland and Labrador’s government in late 2024, which included fake citations. The report was reissued after corrections, and the government updated its procurement policies to require disclosure of AI use and risk assessments in future contracts.
Europe’s cybersecurity agency ENISA admitted in 2025 that two of its threat reports contained multiple AI hallucinated sources. Out of 492 footnotes, 26 were incorrect, raising concerns about the unchecked use of AI in official cybersecurity publications.
Why It Matters
These incidents underscore the growing risks of integrating AI into government processes without sufficient verification measures. They raise questions about accountability, the reliability of AI-generated data, and the potential for reputational damage. As governments increasingly rely on AI for policy drafting, research, and reporting, the need for rigorous human oversight becomes critical to prevent misinformation and maintain public trust.
AI verification tools for government documents
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Background
Over the past two years, AI hallucinations have increasingly appeared in official documents across various countries, often due to the use of generative AI tools like ChatGPT. While AI can assist in drafting and research, these incidents reveal the dangers of unverified AI outputs, especially in sensitive or authoritative contexts. The South African policy incident marked a notable first in government withdrawal, but similar errors have appeared globally, prompting calls for stricter AI governance and verification protocols.
“There will be consequence management for those responsible for drafting and quality assurance.”
— Solly Malatsi, South Africa’s Minister of Communications and Digital Technologies
“ENISA let AI touch the one layer it should never touch unguarded: the truth layer.”
— Chiara Gallese, AI law and data ethics researcher
AI-Powered Software Testing: Volume 1: Foundational Patterns and Principles for Architects and Technical Leads
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What Remains Unclear
It remains unclear how widespread AI hallucinations will become in official government documents in the future, or whether new regulations will effectively prevent such errors. The extent of internal safeguards and verification protocols across different agencies is still emerging, and the long-term impact on public trust is uncertain.
PERPLEXITY AI FOR BEGINNERS: The Complete 2026 Guide to AI-Powered Search, Deep Research, and Smarter Productivity For Students, Professionals, and Everyday Users
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What’s Next
Governments are expected to implement stricter AI verification procedures and update procurement policies to include AI risk assessments. Ongoing monitoring and audits of AI-generated content in official documents are likely to increase, aiming to prevent future incidents and restore public confidence.
AI Programming Made Practical: A Step-by-Step Guide to Building AI-Powered Applications, Writing Better Code Faster, and Using Modern AI Tools with Confidence
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
How common are AI hallucinations in government documents?
While incidents are still relatively rare, recent cases indicate that AI hallucinations are becoming more frequent as governments experiment with AI tools without thorough verification processes.
What measures are governments taking to prevent future errors?
Many are updating procurement policies, requiring disclosure of AI use, and establishing verification protocols to ensure human oversight of AI-generated content.
Could AI hallucinations cause serious harm to government decision-making?
Yes, if unchecked, false data or citations could influence policy decisions, public communication, and international relations, emphasizing the need for rigorous oversight.
Are there legal consequences for AI-generated errors in official documents?
Legal consequences are not yet clearly defined, but governments are beginning to hold responsible parties accountable for oversight failures, as seen in South Africa’s case.
