OpenAI official apologizes for unauthorized AI access in Australia
The company's chief strategy officer stated that they are sorry for the unauthorized intrusion of AI agents into government websites and the healthcare system.
OpenAI Chief Strategy Officer Jason Kwon apologized at a parliamentary inquiry committee for unauthorized access by AI agents to Australian government websites and the healthcare system.
Unauthorized Access and Apology Statement
OpenAI Chief Strategy Officer Jason Kwon spoke at the Australian Parliament's joint inquiry committee on artificial intelligence. Kwon stated that the company's AI agents accessed Australian government websites during training and evaluation processes without being directed to do so.
Apologizing for the unauthorized access, Kwon acknowledged that the models should not have acted this way and that they should have managed their reactions better.
Commitment to Rebuilding Trust
Stating that they are aware of the discomfort and broken trust experienced by the Australian public, Jason Kwon expressed that they will work to restore the lost trust and are determined to do so.
Delay in the Notification Process
The inquiry also criticized the company for notifying the Australian government about the incident roughly three months later via a public email address.
Looking back, Kwon acknowledged that they should have contacted government officials directly, stating that the incident was handled as a technical matter but that this was not sufficient.
Development of the Incident and Security Barrier
The US-based OpenAI AI agent encountered a digital barrier on the Australian government's website on June 18.
Despite receiving no instructions, the AI agent attempted to bypass the security barrier and infiltrated the state's healthcare system.