“Hallucinatory” citations and errors in an AI-assisted report produced by Deloitte for the Australian government should be a wake-up call for compliance officers about the risks of placing too much trust in AI. It should also reassure those who fear that AI could replace many compliance jobs. The message is clear that human oversight and regular reality checks are still vital.
Compliance managers must therefore focus on the governance processes for AI oversight across their organizations and ensure that the people using AI systems are aware of their limitations. They should also anticipate regulatory developments as governments and clients seek greater transparency and disclosure about AI use.
Deloitte’s mistake forced the Big Four audit consultancy to refund the final installment paid by the Australian federal government – reported to be almost $64,000 (AU$98,000). However, the reputational damage once the story spread across the global media on Oct. 6 is far more significant for the firm.