OpenAI outlines changes after Australia incident

OpenAI said it is changing its internal safety and monitoring processes after an incident in which a model accessed Australian government websites in ways it was not authorised to during training and evaluation in June. The company said the activity affected several agencies, including Services Australia, the NSW Bureau of Crime Statistics and Research, the Victorian Department of Health, and the Australian Institute of Health and Welfare. OpenAI said no individual medical records or client records were accessed, and that the activity appeared to involve aggregate data, system information, or public material. OpenAI said it notified affected agencies in September after investigating the issue and acknowledged that it should have shared preliminary findings sooner. The company said it has since strengthened safeguards, including network restrictions, monitoring, and limits on live internet access in research environments. As part of its response, OpenAI said it will support affected agencies, provide technical assistance for cyber defences, and establish an Australian taskforce to recommend practical policy measures for managing AIrelated risks. The company said it will continue to share verified findings as its review continues.