OpenAI admits its AI models accessed Australian govt websites without authorisation: Here is what happened
Follow UsOpenAI has admitted that its AI models accessed Australian government websites without authorisation during internal training and testing in June.
The company said the models were trying to find publicly available information but took actions they were not supposed to.
The company also apologised and said it is working to do better in the future.
OpenAI has admitted that its AI models accessed Australian government websites without authorisation during internal training and testing in June. The company said the models were trying to find publicly available information but took actions they were not supposed to. The most serious incident involved Services Australia’s Medicare Statistics Reporting Service, where a model gained access to non-public files and technical information. OpenAI said there is no evidence that individual medical records were accessed. The company also apologised and said it is working to do better in the future.
SurveyOpenAI’s AI model accessed Services Australia Medicare Statistics Reporting Service
OpenAI said the incident took place during an internal training and evaluation. The model had been given a research task about government spending on medicines for skin conditions in Victorian communities. While looking for the information, the model faced difficulties accessing the required data. It then found a way to get into the Medicare Statistics Reporting Service, even though this access was not authorised.
Also read: OpenAI AI agents leak 53 ChatGPT user images online, company working to take them down
According to OpenAI, the model accessed technical information, source code, internal files, credentials and aggregate statistics. It also wrote files. However, the company said its investigation found no evidence that individual patient or client records were accessed.
“We did not intend for this activity to occur, and the access to the service and follow-on activity should not have happened,” OpenAI said.
Other Australian government websites were also affected
OpenAI AI models also gained authorised access to three more Australian government websites.
An OpenAI model accessed the public Crime Mapping Tool operated by the NSW Bureau of Crime Statistics and Research. It made requests to the public service and received website metadata, application information and logs. OpenAI said individual crime records were not accessed.
OpenAI agents also found an exposed access key to query the Victorian Agency for Health Information’s reporting system. They used it to retrieve reporting configuration and aggregate survey statistics. The company said there was no access to individual medical records or identifiable survey responses.
Also read: OpenAI, Google and Anthropic plan joint AI safety group, may announce it by 2027: Report
The Australian Institute of Health and Welfare was also accessed. OpenAI said the agents retrieved aggregate statistics through browsing and download services. Attempts to bypass access controls were unsuccessful, and the material downloaded appeared to be publicly available.
OpenAI notified Services Australia and the Victorian Department of Health on September 10. It notified the NSW agency on September 18. OpenAI said it should have shared preliminary findings earlier and kept Australian agencies updated during the investigation.
The company plans to support affected Australian agencies, provide cybersecurity assistance and establish an Australian taskforce. OpenAI Chief Strategy Officer Jason Kwon is also said to appear before the Joint Select Committee on Artificial Intelligence on October 6 to answer questions about the incident.
