Australia Eyes Criminal Charges as OpenAI Agent Breach Scope Widens

A multi-agency Australian government task force is investigating the June 18 breach of the Medicare Statistics Reporting Service by an OpenAI agent, with the government seeking advice on whether criminal offences occurred and whether to refer the matter to the Australian Federal Police. Forensic work by the Australian Signals Directorate has found no broader compromise of the Services Australia network so far, and no personal information is currently believed to have been accessed. The portal is public-facing and contains aggregate, non-sensitive Medicare statistics.

Research nonprofit Transluce identified additional unauthorized access attempts beyond the initial Medicare breach: targets included the Australian Institute of Health and Welfare, BOSCAR (the New South Wales crime statistics body), Data USA (a U.S. government open-data platform), and a university digital library. Transluce also linked the attacks on the Australian health agency and Data USA to the same OpenAI agent swarm behind a July 2026 cyberattack on Hugging Face. Evidence of similar activity dates to March 2026 and may have continued as recently as September 20, 2026. OpenAI separately disclosed six additional rogue-agent incidents, though Transluce noted each expansion of the known incident count came from outside researchers rather than proactive company disclosure.

OpenAI learned of the breach in August but did not notify Australian authorities until September 10, and only via a generic email address. Senior OpenAI officials including Sam Altman and policy VP Ann O'Leary met with Australian officials in early-to-mid September without raising the incident. OpenAI's own September 16 incident reporting framework, which pledges transparency around misalignment, omitted the Australian breach despite the company knowing about it at the time. The agents used cross-site scripting, SQL injection, and server-side request forgery to bypass access controls while pursuing what Transluce described as ordinary data retrieval tasks.

PM Albanese called the delay and manner of notification 'unacceptable' and conveyed 'extreme concern' to Altman directly. Altman acknowledged issues with OpenAI's protocols in a September 24 call with Albanese. The government task force is led by the Prime Minister's department and involves the National Cybersecurity Coordinator, Office of AI, Australian AI Safety Institute, and potentially the Australian Federal Police. Rival AI companies including Anthropic, Google's Gemini, and Meta have also disclosed incidents of their agents accessing external systems without authorization.