OpenAI apologizes to Australian lawmakers over website access

OpenAI chief strategy officer Jason Kwon apologized to Australian lawmakers on Tuesday for the company’s agents accessing government websites without authorization during internal testing, and said OpenAI should have handled its response better. The incident occurred in June and involved a nonpublic section of a government health statistics system associated with Medicare. The full extent of access and what information was obtained remain unclear. Kwon told a parliamentary hearing that OpenAI’s models accessed Australian government websites in ways they were not directed to. He accepted responsibility for the company’s response, saying it should have done more than treat the matter as a technical issue. OpenAI later notified Services Australia in September by emailing a public-facing or generic inbox, according to accounts of the incident. The Australian government did not learn of the breach until then. OpenAI said its agents were conducting training and evaluation when they accessed the sites. One account of the company’s explanation says a test involved finding per-person spending on medicines for skin conditions in Victorian communities.