Artificial intelligence companies could face criminal charges for failing to prevent harmful conduct as part of reforms triggered by the breach of a government website.

Criminal penalities, mandatory reporting and a boost to government cyber security defences were flagged as the federal government unveiled the findings of a review following an OpenAI agent gaining unauthorised access to a Medicare statistics website.

WATCH ABOVE: Anthony Albanese unveils new AI safety protections

Know the news with the 7NEWS app: Download today

On Tuesday, November 10, Prime Minister Anthony Albanese announced the findings of the review into the incident, along with the unveiling of an AI safety regulator.

Albanese said the government wanted to embrace the economic opportunities presented by artificial intelligence while at the same time ensuring adequate safeguards were in place.

“Australia is a first mover in this space, and we want to make sure that we reap the potential benefits of AI,” he told reporters in Canberra.

“But we need to ensure that we get it right, and social licence is absolutely critical.”

Anthony Albanese has flagged mandatory AI incident reporting to prevent future cyber incidents. Credit: AAP

Albanese did not say whether OpenAI’s breach constituted a breach of laws, but said measures were needed to prevent future incidents.

“The point is law needs to get ahead of, and in some cases, make sure that it catches up with what is a very fast-changing world,” he said

Be the first to know:

“There are obviously some national security implications and we won’t be assisting those who want to undermine.”

While the 18 recommendations have been handed down, the full report has yet to be publicly released.

“Our task is to make sure that we shape AI rather than AI shape us,” Albanese said.

“It will have a drastic impact on the way that work occurs.

“AI offers Australia and the world generational opportunities if we seize and shape them properly and that is exactly what we’re doing here.”

Calls for criminal penalties for AI agents allowed to go rogue

OpenAI’s technology accessed a Medicare statistics website in June, but the firm did not inform federal authorities about the incident until September. Credit: AP

Among the most significant recommendations is the introduction of mandatory, time-bound reporting obligations for frontier AI developers when serious incidents occur.

The review also calls for stronger information sharing between government agencies, closer cooperation with international AI developers and consideration of new criminal penalties for companies that fail to prevent harmful conduct by their AI systems.

Under proposed national standards for AI, which will be introduced to parliament, AI companies would need to allow for independent evaluations of their models

Security controls would also need to be maintained across all stages of development, with safety and security frameworks needing to be made public

The Attorney-General’s Department has been asked to examine potential legal reforms, including penalties for serious regulatory breaches and the creation of a possible “failure to prevent” offence targeting companies whose AI models cause harm.

An OpenAI model gained access to Medicare statistics in June, though the Australian Government was not notified until months later. Credit: AAP

The review found Australia must improve its preparedness for AI-related cyber incidents by strengthening reporting pathways, improving agency coordination and bolstering cyber security defences across government systems.

Federal agencies will also be encouraged to reduce their online exposure by consolidating digital systems and disconnecting unnecessary internet-facing infrastructure.

The government said agencies would be directed to improve staff understanding of AI risks while accelerating adoption of stronger cyber security practices.

The reforms are expected to form part of a broader effort to establish Australia as a leader in AI governance as governments around the world grapple with the rapid rise of increasingly powerful artificial intelligence systems.

With AAP