Deloitte to Partially Refund Australia Amid Controversy Over AI-Generated Errors in Report

MELBOURNE, Australia – Deloitte Australia is set to partially refund the Australian government after a report they produced contained numerous errors, including a fake quote from a court judgment and references to non-existent research.

The report was delivered to the Department of Employment and Workplace Relations in July, but concerns about its accuracy surfaced when Chris Rudge, a researcher at Sydney University, alerted the media. He pointed out that the report was riddled with fabricated references, leading to a revised version being published recently.

In a statement, the government confirmed that Deloitte looked over the 237-page document and acknowledged some footnotes and references were indeed wrong. The department noted that Deloitte agreed to repay the final payment under their contract, although the exact refund amount will only be revealed later.

When asked about the inaccuracies, Deloitte stated that the matter had been resolved with the client but did not comment on whether AI was responsible for the errors. The tendency for generative AI to create false information is known as “hallucination.”

The report focused on the use of automated penalties in Australia’s welfare system. Despite the errors, the department claimed the overall substance of the report remained untouched, and the recommendations were unchanged. However, the revised report now includes a note indicating that a generative AI tool, Azure OpenAI, was used in its creation.

Rudge discovered as many as 20 errors in the original report. One glaring issue was a statement about a non-existent book supposedly written by a law professor, which Rudge recognized immediately as either an AI fabrication or an extreme oversight. He expressed that misquoting a judge is a serious problem, especially when the report served as an audit of the department’s legal compliance.

Senator Barbara Pocock, representing the Australian Greens party, criticized Deloitte for their misuse of AI. She suggested the full AUD 440,000 ($290,000) should be refunded. Pocock argued that such fundamental errors would lead to significant consequences for a university student.

This incident highlights the increasing concerns over AI’s reliability. A recent survey showed that 72% of experts worry about AI systems creating false information, emphasizing the need for more stringent oversight when using these technologies in critical areas. As AI continues to evolve, it is crucial for companies to approach its integration with caution and responsibility, ensuring accuracy and accountability in their outputs.

For more on this topic, you can check out the Australian Broadcasting Corporation’s coverage of the issue here.

Source link

Artificial intelligence, Business, News media, General news, Courts, Technology, World news, Article, 126281611