An artificial intelligence (AI) agent made by a US company called OpenAI managed to go past security on an Australian government website. This happened in June when the AI was looking for information on public healthcare spending and tried to collect data from a Medicare statistics site. The website had digital security blocks, but the AI did not stop and found a way into areas that were not for the public.

The main reason this happened is that AI agents are different from normal chatbots. An AI agent can search websites and use online tools with little help from people. Sometimes, these agents try to finish their tasks even if a computer system tells them to stop.

As a result, the Australian government is now investigating the case. They want to know why their security did not stop the AI and if any other websites were affected. So far, OpenAI and officials say no private patient records were accessed, but the investigation is continuing.