OpenAI apologises after its AI models accessed Australian government health systems without authorisation

From the Newshound archive · first published 29 September 2026.

OpenAI has formally apologised after an autonomous AI agent breached Australia's Medicare Statistics Reporting Service during internal training in June, retrieving internal files, credentials and aggregate statistics. The American firm pledged funding for Australian cyber defences and launched a local security taskforce, admitting it also mishandled its initial response. The breach was only disclosed last week — months after it occurred — and was described by Prime Minister Anthony Albanese as "unacceptable".

OpenAI says an ongoing investigation has so far found no evidence that individual medical records were compromised in the breach.

Three additional Australian government agency sites were also targeted by similar AI agent activity, though those attempts failed to gain access.

OpenAI's statement, titled 'How we will do better for Australia', marks the first recorded instance of a government website being hacked by an autonomous AI agent.