Artificial Intelligence

Deloitte scandal: must compensate Australian government for report full of AI-generated errors

The document, produced with the help of a large language model, contained invented quotations and references. The company will have to return part of the fee and face image damage

by Massimo De Laurentiis

(Imagoeconomica)

2' min read

Translated by AI
Versione italiana

2' min read

Translated by AI
Versione italiana

By now everyone knows: generative artificial intelligence errors, so-called hallucinations, are an integral part of this technology and cannot be eliminated altogether.

Yet Deloitte, one of the world's largest consulting firms, will have to compensate the Australian government for delivering a report full of inaccuracies produced by AI.

Loading...

The report, which cost Canberra's coffers about $290,000, was published in July on the Department of Employment and Labour Relations website. Only later did Chris Rudge, a researcher at the University of Sydney, realise that the text contained incorrect references and citations to non-existent books.

The case was reported to the media and on 3 October, the department released a corrected version of the report. In the new document, Deloitte admits that Azure OpenAI GPT-4o, a large language model, was used at some stages of production.

"The report has been updated to correct citations and bibliography entries that contained errors," reads the revised text. "The updates made in no way affect the substantive content, conclusions and recommendations of the report."

In addition to correcting the report, the company will have to repay part of the amount paid by the Australian government.

This episode is reminiscent of some past cases in which blind trust in artificial intelligence has led to glaring errors. In 2023 in New York, lawyer Steven A. Schwartz cited a series of non-existent legal rulings, completely invented by ChatGPT. Another case concerns the airline Air Canada, which was forced to compensate a customer who had been given incorrect information by a customer service chatbot.

The image damage for Deloitte is likely to be even greater, because it comes at a time when the risks associated with the use of AI are now well known and more regulated than in the past. Moreover, the consulting agencies that are part of the Big Four, including the British multinational, have invested billions of euros in recent years in the development of AI in order to make their activities faster and more efficient.

Deloitte itself has a section of the company dedicated to AI, which includes training services. Through the Deloitte AI Institute, the company aims to drive the development and adoption of responsible and trustworthy AI. On the institute's Italian website, the company states that it guarantees solutions that are based on 'in-depth knowledge of the sector' and insists on security, impartiality and transparency.

Statements that clash with a gaffe described as a mistake 'for which a first-year university student would have serious problems', by Australian Senator Barbara Pocock.

This scandal, moreover, comes after the UK's accounting regulator accused the Big Four of not giving due consideration to the impact of AI on the quality of their audits last June.

Copyright reserved ©
Loading...

Brand connect

Loading...

Newsletter

Notizie e approfondimenti sugli avvenimenti politici, economici e finanziari.

Iscriviti