VideoRecent global cyber incidents and technical failures have disrupted essential services including Microsoft, Apple, Medicare and airline systems, with Australia experiencing a 20% increase in cyber incidents over the past year.

A new report into OpenAI’s hack of Australia’s Medicare website has found prosecution gaps, an inadequate disclosure timelines by the tech giant and government escalation failures.

The Australian government has released its full report into the incident where a rogue AI agent from the developer of ChatGPT had acted independently to gain unauthorised access to Services Australia on June 18.

Anthony Albanese had released the 18 recommendations of a rapid review earlier this month but on Friday provided the complete investigation including an overview of the incident and the government’s incident management.

OpenAI claimed they identified the breach in August but it took until September 10, nearly three months after the hack, for the firm to alert the department via email.

The report claimed the AI agents had been looking for data into medicine spending for skin conditions in Victorian communities.

Read more...

While the Australian Government said no patient records had been accessed and the “non-public” information it had hacked wasn’t sensitive, the incident has prompted an overhaul of cybersafety.

The new report showed that OpenAI initially treated the event primarily as a technical vulnerability rather than a cyber incident which had led to an unusual disclosure process.

The tech firm had contacted Service Australia’s public facing address public.disclosure@servicesaustralia.gov.au which is a dedicated contact to report cyber security incidents.

Open AI had found the breach on August 11 following a review prompted by a high-profile July incident involving Hugging Face.

Despite company representatives meeting with Australian politicians Richard Marles and Andrew Charlton on September 1, Open AI took until September to notify Services Australia.

One of OpenAI’s most senior executives, Jason Kwon, had travelled to Australia in October to deliver an in-person apology to an Australian parliamentary inquiry for breaches.

But even after the tech firm eventually contacted Australia, the report shows that there was a multi-day delay in escalation through the government, which lacked a clear whole-of-government approach.

Their email to Services Australia’s dedicated cyber incident reporting line took the agency five days to escalate to the Australian Signals Directorate.

It was also only reported to the National Cyber Security Coordinator more than a week later on September 19.

At the time of the breach, it was mandatory for government agencies to report significant security incidents to the Australian Signals Directorate and Home Affairs, however, “cyber vulnerabilities” was only “encouraged”.

In the wake of the hack and with the acceleration of AI, however, it noted the distinction between a vulnerability and an incident may become less clear.

It has since prompted Home Affairs to consider a new requirement under the guidance, including appropriate escalation timeframes.

All agencies have since been told to review cyber incident reporting and escalation procedures and encouraged to use AI to strengthen defence.

The report said departments should take lessons from Australian Institute of Health and Welfare’s “success” of stopping AI agents in the same attack from accessing unauthorised material through “strong cyber fundamentals”.

The government has invested around $160 million for cyber security uplift for Services Australia, which has also updated its procedures so disclosures are forwarded to its 24/7 Cyber Security Operations Centre.

The report also pointed to gaps in criminal prosecution for such events and insisted the breach couldn’t be referred to the Australian Federal Police under current laws because human intent or knowledge was absent during the automated incident.

Mr Albanese had revealed earlier this month that Australia was considering tougher laws to criminally prosecute companies responsible for harmful AI breaches.

It also found that Australia needs friends in its fight for safety, insisting that regulation in isolation would be challenging.

“Significant divergence from international approaches could create enforcement difficulties and reduce the practical reach of Australian rules,” it stated.

“Particularly where firms are able to shift development or deployment activities to jurisdictions with less stringent requirements.

“This reinforces the importance of international cooperation to support effective regulation and promote shared expectations regarding AI safety and accountability.”

The PM also confirmed that the government had officially accepted all 18 recommendations put forward by the rapid review.

DOWNLOAD THE NIGHTLY APP FOR THE LATEST BREAKING NEWS

Get the latest news from thewest.com.au in your inbox.

Sign up for our emails