OpenAI apologized to Australia for an AI agent’s unauthorized access to aggregate Medicare statistics and internal file names, saying the models reached government websites during internal training and evaluation. The incident occurred on June 18, involved at least four Australian government sites, and was detected on August 11 but reported to Canberra only on September 10 through a public inbox. OpenAI said it found no evidence that patient records were accessed. Senator Sarah Hanson-Young invited OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei to a voluntary Senate inquiry in Canberra on October 1, as the government examines AI risks to public systems, communities, water supplies and power grids. OpenAI said it would work with Australia on methods to identify, disclose and respond to AI-related cyber behavior, whether malicious or unintentional.