OpenAI apologizes for hacking Australian government websites

OpenAI has apologised to Australia for an incident in June 2026 when its AI agent hacked into government websites and a closed statistical database of the Medicare healthcare system. In a statement, OpenAI said it would correct the mistakes and provide funds to strengthen cyber defenses in the country.

The company intends to establish an expert working group in Australia with the participation of local experts. By the end of 2026, the group will develop recommendations for the interaction of AI developers with the government. OpenAI has also committed to reviewing its security protocols. OpenAI will provide technical assistance and funding from the global Daybreak for Frontline Defenders fund to protect critical infrastructure from AI threats.

During a trial on June 18, OpenAI’s research team tasked the AI agent with finding data on government spending on drugs in Australia. AI agents went beyond the task, hacking several closed databases, including the portal Services Australia, as well as health ministries of several states. AI agents downloaded the data and recorded it to the system’s internal server. OpenAI only learned about the unauthorized activities of its AI agents on August 11 during an internal audit. The company notified the Australian government of the incident on September 10 by sending an email to report vulnerabilities. The Australian government announced its intention to tighten the regulation of AI systems in connection with the incident. OpenAI head Sam Altman will speak in the Australian Senate as part of the Greens investigation.

Alena Miklashevskaya