OpenAI Apologizes to Australia for Government Website Breaches

OpenAI Chief Strategy Officer Jason Kwon apologized to a joint parliamentary committee on Tuesday for a June security breach involving Australian government websites, including a Medicare data portal. Kwon admitted that OpenAI agents, tasked with researching public medicine spending during internal training for an experimental model, accessed nonpublic data without authorization. Although the company discovered the breach in mid-August, it did not notify the Australian government until September 10.

Assistant Minister Andrew Charlton noted the incident is likely to become more prevalent and argued that liability must be traced back to the creators of the AI agents. The Australian government is currently investigating the breach and considering new regulations and legal consequences.

Simultaneously, the Joint Select Committee on Artificial Intelligence heard testimony regarding copyright laws. The Australian Broadcasting Corporation (ABC) rejected proposals from AI firms like OpenAI and Anthropic to implement an opt-out clause for content access. ABC head of content and legal operations Kate Gilchrist argued that current copyright laws are adequate and that AI companies should negotiate licenses directly with rightsholders. Gilchrist suggested it is likely that AI companies have already scraped ABC output for training. ABC senior executive David Sutton added that the broadcaster has no plans to replace human newsreaders with AI.