Report from ABC News
In Brief – Consulting firm Deloitte Australia will refund part of the AU$440,000 ($290,000) it was paid by the Australian Government’s Department of Employment and Workplace Relations for a report assessing the agency’s use of automated welfare penalties after the document was found to include at least 20 fabrications and inaccuracies that appear to have been generated by an AI system. The initial 237-page report, which was posted on the agency’s website in July, included numerous false citations, references to non-existent academic papers, and a fabricated quote attributed to a federal judge who is a Sydney University professor of public and constitutional law. The errors were caught by a Sidney University researcher who was tipped off by the fabricated quote linked to a book the Sidney University law professor had not written. Deloitte confirmed that “some footnotes and references were incorrect” and that the “matter has been resolved directly with the client.” The company initially did not respond when asked if the errors were generated by AI, but later confirmed they were. The revised report, posted in October, disclosed that Microsoft’s Azure OpenAI system had been used in its preparation. The department said the core findings and recommendations remained unchanged.
Context – In July, a US district judge in New Jersey withdrew his decision in a biopharma securities case after lawyers pointed out that his opinion contained numerous errors, including made-up quotes and misstated case outcomes, that were very likely the product of AI. Chatbot “hallucinations” in legal briefs first made news in mid-2023, less than a year after ChatGPT’s release, with two lawyers being sanctioned in US District Court. Chief Justice John Roberts led off his 2023 Annual Report on the Federal Judiciary by discussing AI technology in the legal system, warned about hallucinations, and urged responsible use. The Judicial Office of the UK Ministry of Justice released similar AI guidance. Legal world hallucinations keep popping up. Of course, everyone knows that they need to check all AI work, but everyone also knows they should not drink and drive.