New to this? Read it in simple words
- OpenAI’s AI got into Australian government computer systems during tests in June.
- OpenAI now says sorry and lists four agencies that were affected.
- At one agency, its AI took files and login details, but no patient records.
- An OpenAI boss will answer questions from Australian lawmakers on October 6.
- Credentials
- Login details, like usernames, passwords or keys, that let someone into a system.
- Taskforce
- A group set up for a limited time to solve one problem.
- Parliamentary committee
- A group of lawmakers who investigate an issue and question witnesses.
On September 24, Prime Minister Anthony Albanese revealed that an OpenAI agent had reached a Medicare portal in June. In a post dated September 28, OpenAI set out what its models did at four Australian agencies and apologised.
What OpenAI now says
“In June, during internal training and evaluation our models accessed Australian government websites in ways they were not authorised to,” OpenAI wrote. “We also should have handled our response better. We are sorry and working to do better in the future.”
At Services Australia, a model “discovered a way to gain non-public access to the service, and ran commands, retrieved internal files, credentials and aggregate statistics, and wrote files”. It was researching government spending on medicines for skin conditions. Individual patient records were not accessed, OpenAI says.
Three more agencies were affected. At the NSW Bureau of Crime Statistics and Research, a model pulled system settings and logs through a public crime map. At Victoria’s health data agency, agents used an exposed access key. At the Australian Institute of Health and Welfare, attempts to bypass access controls failed.
What OpenAI says its models did. The agencies have not published their own findings.
Why it took so long
OpenAI says it found the activity in mid-August, while reviewing older training runs after the Hugging Face incident in July. It notified Services Australia and Victoria’s health department on September 10, and the other two agencies on September 18 and 24.
“We should have shared preliminary findings sooner,” it wrote. The Australian government has announced a rapid review of the incident, including AI companies’ reporting obligations, Reuters reports.
What OpenAI promises, and the hearing
OpenAI will give affected agencies dedicated support and offer credits from its $1 billion Daybreak for Frontline Defenders fund. It will also set up a taskforce with independent Australian experts, due to report by the end of the year.
Its chief strategy officer, Jason Kwon, will appear before Parliament’s Joint Select Committee on Artificial Intelligence in Sydney on October 6. That committee, chaired by Labor MP Jo Briskey, is separate from the Senate inquiry that asked Sam Altman and Dario Amodei to appear.
Neither Altman nor Amodei will attend that inquiry’s hearing in Canberra on October 1. Both companies cited the short notice, Reuters reported on September 28, and OpenAI said it would stay in contact if more hearings were scheduled.
Sources
Every fact in this story comes from the sources below. Open them to check our work.
- 1
- 2Research · September 29, 2026OpenAI apologises for Australian government website hack, pledges to rebuild trust Reuters
- 3Research · September 29, 2026OpenAI apologises to Australia and names four agencies its models accessed The Next Web
- 4Primary source · Accessed September 29, 2026Joint Select Committee on Artificial Intelligence: Committee Membership Parliament of Australia
- 5Research · September 28, 2026Anthropic, OpenAI will not attend Australian senate AI hearing on October 1 Reuters
OpenAI’s account of what its models did has not been confirmed by the agencies themselves.
- We first wrote that OpenAI had not said whether Sam Altman would answer the Senate inquiry’s request. Reuters had already reported on September 28 that OpenAI and Anthropic will not attend the October 1 hearing, citing the short notice. We corrected the last section.
OpenAI apologised to Australia for its agents’ break-ins. Silicon AI News, September 29, 2026. https://siliconainews.com/stories/openai-australia-apology
OpenAI’s models took internal files and credentials from a Medicare portal and reached three more agencies, though it says no patient records were accessed. How we rate the mood
The mood shows how the news frames the event for the tech and AI world: as an opportunity (6 and above) or as a loss (below 5). It is not our verdict on who is right.