Ethics fears over Queensland departments’ AI policies
A Queensland Audit Office report has found limited visibility in AI use across state government could be posing risks.
The Department of Transport could be breaching Queenslanders’ privacy, a report has found, over its failure to identify ethical risks in AI image-recognition technology.
A Queensland Audit Office report into a departments, including The Department of Customer Services, Open Data and Small and Family Business (CDSB) and the Department of Transport and Main Roads (TMR), found there was limited oversight into how AI is implemented and monitored.
The report found TMR had not yet introduced comprehensive arrangements to monitor the ethical risk of AI. TMR currently employs the use of AI systems including the Mobile Phone and Seatbelt technology (MPST) image-recognition and QChat systems.
TMR has evaluated ethical risks for both systems, which were implemented before CDSB’s AI governance policy, which has been in place for 12 months.
The QAO recommended TMR perform a full ethical risk assessment to determine how to best improve their governance and mitigation strategies. The report also advised TMR improves AI oversight and staff capability to use these systems responsibly.
You might like
The audit report states that TMR needs to do more department-wide beyond its initial analysis to manage the ethical risks of AI use in a structured manner.
Meanwhile, the report found CDSB had an effective AI use policy in place, requiring specific governance and supported to assess ethical risks of AI.
However, the report found CDSB has limited visibility on AI use and its possible risks across the QLD Government. The audit stated that CDSB’s management of AI risk is not progressing as fast as AI use is increasing and that its AI policies need strengthening.
QAO’s report said CDSB needs to better understand the Queensland public sector’s use of AI, by focusing on continual improvement of AI policies and risk monitoring.
The report also suggested strengthening assessment tools and to support monitoring of AI assistant systems like QChat to ensure the use of AI is safe and reliable.
