Delegating work to generative AI carries a hidden risk of fabricated facts and declining critical thinking
The cases involving the South African government and Deloitte showed that AI can invent citations and authors. A survey of 48 000 people found that 66 % of them use AI regularly, but 61 % have no training and the main motivation is fear of falling behind, not quality.
The South African government had to withdraw the Draft National AI Policy this year after reviewers found that many citations referred to nonexistent articles and authors – errors introduced into the document by generative AI. Similarly, last year Deloitte refunded part of its fee to the Australian government after admitting that it had used generative AI to prepare a commissioned report, which inserted fabricated citations and invented quotations from sources that do not exist into the text. The report cost Australian taxpayers 440 000 Australian dollars.
According to a 2025 study by University of Melbourne and KPMG, which surveyed over 48 000 respondents in 47 countries, 66 % of people use AI regularly and more than half believe that AI improves their performance. The study states that the main motivation for adopting AI is the fear of falling behind others (48 % of respondents), which leads users to prioritize delivery speed over the quality of the result. At the same time, 61 % of respondents said they had received no training in using AI, and 60 % described inappropriate, permissive or nontransparent use of AI in the workplace.
A separate study by the nonprofit research organization Centre of AI Safety in San Francisco from July 2026 found that even the best-performing AI agents failed to complete approximately 85 % of projects to a standard acceptable for commercially commissioned work.
The author of the article (a cybersecurity expert) argues that relying on polished AI outputs without understanding the process by which AI arrived at them weakens the ability to critically assess facts and distinguish truth from half-truths. As countermeasures, the author proposes realistic time frames for submitting work, stepping away from AI tools during the brainstorming phase, and building a culture that critically examines the logic, structure and sources of all outputs, regardless of whether they were created by a human or AI.
Why it matters
Documented cases show tangible financial and reputational damage when unchecked AI outputs make their way into official documents – from the withdrawal of a government policy to the refund of hundreds of thousands of dollars from a public contract. At the same time, the low level of user training (61 % without training) and motivation based on fear of falling behind, rather than quality, suggest that similar risks are likely recurring in companies and institutions unnoticed until a visible error emerges.
Two audiences, two different impacts
What this means
For individuals
Anyone who uses AI outputs without verifying the underlying source material will not recognize the risk of fabricated facts or citations until it causes a problem – and without training (which 61 % of users lack, according to the cited survey), these blind spots are harder to recognize.
For a business
Unchecked use of generative AI to prepare reports and documents can lead to tangible financial and reputational losses, as demonstrated by the fee refund by Deloitte and the withdrawal of a government policy in South Africa over fabricated citations.
Risks and complianceCheck the original
Event sources
only one source so far · 1 publisher, 1 independent. We count feeds from the same owner only once.