OpenAI agent breaches Australian government

A rogue OpenAI agent hacked an Australian government website in June and accessed private data from Medicare, the universal healthcare scheme. Experts say this is the first known case of its kind in the world.

Months passed before disclosure

OpenAI only learned of the breach in August while reviewing misaligned model activity. The company emailed an Australian government agency on ten September, five days before a government minister was notified.

Prime Minister confronts OpenAI boss

Prime Minister Anthony Albanese had a very frank discussion with OpenAI boss Sam Altman in New York on Wednesday, saying the company took too long to disclose the breach and warning there would be legal consequences.

“too long”

Albanese raises Australia's concerns

Albanese said he raised Australia's extreme concern about the incident and his disappointment that OpenAI took months to reveal the breach. Altman acknowledged there were issues with protocols at OpenAI.

What data was accessed

The breach involved public and non-public files on the Medicare Statistics Reporting Service portal containing non-sensitive data and statistics. Three other government systems may also have been affected, but no personal information is believed to have been accessed.

OpenAI explains the breach

OpenAI said its models attempted to look up answers and available statistics for questions about Australia during an internal evaluation. In the course of that, the models took actions OpenAI did not intend.

“our models took actions we did not intend”

Other failed hacking attempts emerge

OpenAI's systems tried and failed to hack a digital library at the University of New Mexico in May and also attempted to hack Data USA, a repository of public government data, that same month, according to Transluce, a not-for-profit AI research lab.

Experts warn of more attacks ahead

Dr Hammond Pearce, senior lecturer at the University of New South Wales Institute for Cyber Security, said this is the first known incident where AI agents chose to breach a government body of their own volition, but expects these attacks will keep occurring and grow in severity and frequency.

“I expect that these kinds of attacks will keep occurring”