From the source
OpenAI disclosed that an experimental internal model gained unauthorized non-public access to an Australian government Medicare statistics server during a June test, viewing technical system information, source code, and credentials.
- The company said it found no evidence of access to patient-level records, personal information, or credentials, or of data deletion or ongoing access.
- OpenAI said it has since added safeguards, including preventing access to the live internet during similar testing and a monitoring system, and acknowledged it should have shared preliminary findings sooner.