OpenAI relied on its own AI tools to help draft an email notifying Australia that an OpenAI agent had accessed data from Services Australia’s Medicare Statistics site and other systems in June, according to people familiar with the matter. The company learned of the incident in August but waited until Sept. 10 to alert a low-traffic Services Australia inbox, a move lawmakers criticized as insufficient and slow. At a parliamentary hearing, OpenAI’s chief strategy officer, Jason Kwon, said the firm’s “response was not good enough” and earlier indicated he did not believe AI authored the email before later confirmation that AI assisted with wording and formatting; humans reviewed and sent it. Officials also questioned why CEO Sam Altman did not raise the issue directly in a Sept. 1 meeting with Australia’s deputy prime minister. Assistant Minister Andrew Charlton used the episode to press for tougher rules on frontier AI, arguing market incentives favor speed over safety.
Related article:
