Queensland AI traffic cameras lack ethical risk assessment
Queensland's AI-powered traffic cameras used to detect mobile phone and seatbelt offences did not undergo required ethical risk assessments, a state audit found. The audit identified concerns about privacy, inadequate human oversight, inaccurate image recognition, and photo handling, with cameras making 208 million assessments and issuing approximately 114,000 fines in 2024.
What happened
The Queensland Audit Office found that the Department of Transport and Main Roads (TMR) did not conduct required ethical risk checks before deploying AI-powered cameras to detect mobile phone and seatbelt offences.
The audit identified multiple risks that were not properly examined, including inaccurate image recognition, handling and storage of driver photos, and inadequate human oversight in the fine-issuing process.
Reported impact
In 2024, the cameras made more than 208 million assessments and issued approximately 114,000 fines, generating more than $137 million in revenue.
The audit warned that potential problems with the system, including privacy concerns and fairness issues in the human review process, could affect drivers receiving fines.
Index analysis
The AI image recognition technology filtered out 98.7 per cent of the 208.4 million assessments as unlikely to show offences, leaving 2.7 million images for human review by an external vendor, followed by a secondary review of 137,000 potential offences by the Queensland Revenue Office.
The Queensland Audit Office found that TMR had not effectively identified the ethical risks of artificial intelligence and had little oversight of how the systems operated, with the state's broader AI governance also found to be falling short.
Governance controls
TMR included some mitigation strategies such as human review of potential offences, but the audit said the 'completeness' and effectiveness of these arrangements needed to be tested and improved.
The audit recommended that TMR perform department-wide ethical risk assessments and create stronger checks and greater transparency around AI use, to which TMR said it had accepted the recommendations and commenced implementation.
Classification
Sources and evidence
Sources not yet published
This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.