Recently, the child protection agency in Victoria, Australia, was forced to issue a ban, requiring staff to cease using generative artificial intelligence (AI) services. This decision was made after an employee used ChatGPT to draft a protection application report and entered a significant amount of personal information, including the name of a high-risk child. This incident drew the attention of the Office of the Victorian Information Commissioner (Ovic), which received a report in December last year and subsequently launched an investigation.
The investigation revealed that the report submitted by the employee involved a young child whose parents were charged with sexual offenses, although these charges did not involve the child directly. Ovic noted in its investigation report that several phrases in the report were clearly generated by ChatGPT and contained language and sentence structures that did not comply with child protection guidelines. For instance, the report described a child's doll, allegedly used by the father for sexual purposes, as a "developmentally appropriate tool to support the child's needs," significantly downplaying the potential risks faced by the child.
Ovic's report also pointed out that inputting personal information into ChatGPT constituted unauthorized disclosure of departmental information. The employee admitted to using ChatGPT to save time and improve efficiency but denied entering personal information. The investigation also found that this employee may have used ChatGPT in 100 cases. Throughout the department, from July to December 2023, nearly 900 employees accessed the ChatGPT website, accounting for nearly 13% of the total staff.
As a result, Ovic issued a series of directives to the department, including blocking the IPs and domains of multiple generative AI websites. ChatGPT, Meta AI, Gemini, and Copilot are all within the scope of the block, and the ban will last for two years. The department has expressed acceptance of the investigation results and will actively implement the relevant directives, and has already terminated the employment of the employee in question. Ovic is open to the future use of generative AI in child protection but emphasizes the need for higher safety standards.
Key Points:
✅ A child protection worker in Victoria entered personal information using ChatGPT, leading to a ban on generative AI use by the agency.
🚫 The investigation found multiple instances of language and inaccurate information in the report that did not meet child protection standards.
🔒 Ovic has required the department to block multiple AI websites, with the ban lasting two years to ensure the safety of children's information.
