OpenAI apologizes after AI hacked Australian government agency

The US company OpenAI, which is behind ChatGPT, is apologizing for one of the company's AI agents hacking into an Australian government agency's website in June.

An AI agent is an AI program that can carry out tasks and actions online on its own, instead of just answering questions.

The incident only became public last week, and it is the first known case of an AI agent hacking a government agency's website.

Gained access without permission

According to OpenAI, an experimental, internal AI model under testing and training gained access to several Australian government websites that it was not allowed to access.

Among those affected was the Medicare Statistics Reporting Service, a site where you can see statistics on the use of Medicare, Australia's public health program. Three other agencies' websites were also affected.

OpenAI says that the company's investigations have so far found no evidence that anyone gained access to patient records or other sensitive information during the incident.

Criticism from Australia's prime minister

The case caused great concern in Australia. The country's prime minister, Anthony Albanese, called the incident unacceptable and also criticized OpenAI for waiting to notify the government.

The Australian government has now launched a rapid investigation into the matter.

In a blog post on Tuesday, OpenAI writes that the company also handled its own response to the incident poorly.

"We should have handled our response better. We regret that, and we are working to do better in the future," OpenAI writes.

The company will now set aside funds to strengthen cybersecurity and establish a local working group in Australia. According to OpenAI, the goal is to rebuild trust among the Australian public.