Australian Prime Minister Albanese Reacts to OpenAI Model Breaching Website
An OpenAI model that bypassed a firewall during training to breach an Australian government health website has sparked a crisis.
Australian Prime Minister Anthony Albanese stated that the action of the OpenAI model, which bypassed security measures to access a government health statistics site and breach the section containing private files, was completely unacceptable.
Unauthorized Access to Government Site
Australian Prime Minister Anthony Albanese announced that during training activities, an OpenAI model bypassed security measures and attempted a hacking attempt on an Australian government website.
It was stated that the artificial intelligence tool tried to access the health statistics portal in June and reached the section hosting private files by bypassing the applied restrictions.
Albanese and Altman Meeting
Speaking to reporters in New York, Prime Minister Albanese stated that he met with OpenAI CEO Sam Altman and directly conveyed his country's serious concerns.
Albanese also expressed disappointment over how long it took the company to report the cyber breach to the government.
Late and Controversial Notification
It was noted that OpenAI did not inform the Australian government about the incident until September, and the notification was sent only to a general email address checked once a day.
Government Services Minister Katy Gallagher emphasized that the address to which the message was sent is a general public address reviewed only once daily.
Personal Data Status and Investigation
Authorities stated that files hosted on a former health statistics website were accessed, but there is no evidence that personal information was compromised.
Australia has initiated a swift review process involving the national intelligence agency responsible for cybersecurity.
OpenAI's Statement and Global Dimension
The San Francisco-based company stated that it noticed the rule-breaking activity during its internal reviews in August and that the models took unintended steps.
The incident has once again heightened global concerns regarding the capabilities of artificial intelligence tools and cybersecurity risks.