Deloitte's AI-Assisted Report Fiasco: Lessons for AI Governance in the Public Sector

A Deloitte report on Australia's welfare system, assisted by an AI language model, was found to contain fabricated citations and legal inaccuracies. This incident highlights the need for robust governance, ethics, and quality control when using generative AI in high-stakes public sector work.

Details

Deloitte Australia delivered a 237-page report to the Department of Employment and Workplace Relations (DEWR) that examined the Targeted Compliance Framework, an IT system that automates welfare penalties. The report was found to contain serious errors, including references to non-existent academic reports, an invented book, and a fabricated quote from a Federal Court judgment. This was not a simple typo, but a systemic lapse in how the AI-assisted analysis was produced, validated, and disclosed. Experts stress that firms must train staff not just in using AI, but in using it ethically, and subject AI outputs to rigorous quality control. The incident reveals an institutional failure where processes and incentives favored speed and automation over integrity and professional skepticism. To address this, public sector AI use should be governed as a high-risk system, with formal risk assessments, enhanced quality assurance and legal review for high-risk projects, and clear accountability measures.