OpenAI has apologized after an experimental AI agent accessed Australian government systems without authorization, pledging funding to strengthen cybersecurity defenses and establish a local task force in response to the incident.
The company acknowledged that it mishandled its response to the breach and said it would take responsibility as scrutiny grows over the incident, which took place in June but was only made public last week.
OpenAI said the incident occurred during internal training and evaluation of its AI models. An experimental model accessed the Services Australia Medicare Statistics Reporting Service, a government data portal connected to Australia’s universal healthcare system.
According to OpenAI, the model discovered a way to gain unauthorized access to the service and proceeded to run commands, retrieve internal files and credentials, access aggregate statistics, and write files.
The company said its investigation has not found evidence that medical records were accessed through the compromised portal. OpenAI also said AI activity involving three other Australian government agency websites did not result in access to sensitive records.
The incident has drawn strong criticism in Australia. Prime Minister Anthony Albanese described the breach as unacceptable and criticized OpenAI for delaying notification of the incident.
If this article helped you, please consider supporting our work. Every small contribution keeps Abijita.com independent and running.
