OpenAI's chief strategy officer Jason Kwon acknowledged during a parliamentary hearing in Sydney that the company's response to a breach of Australian government websites in June was "not good enough." Kwon stated that the breach, which involved an OpenAI agent accessing a private statistics portal containing non-sensitive data from Australia's Medicare system, should not have occurred and that the company should have communicated more effectively with the government. It took weeks for Australia to be notified, with Kwon admitting that contacting government ministers directly was a mistake.
Kwon expressed regret, stating, "We are sorry and we know we have work to do to rebuild trust with the Australian people." OpenAI has since implemented additional precautions in its training environments and is establishing a local taskforce in Australia to better manage risks associated with AI. Kwon noted that training models are now monitored in real time, allowing for quicker alerts to government entities in case of incidents.
Anthropic, another AI company, also participated in the hearing and reported that it found no breaches of Australian government websites during its investigations. The hearings, which will continue until Friday, also addressed concerns from arts and media organizations regarding copyright issues related to AI models. Annabelle Herd, chief executive of the Australian Recording Industry Association, criticized the proposed opt-out model for artists, suggesting it could lead to artists not being compensated for their work.