Factual. Independent. Impartial.
Support AAP with a free or paid subscription
Finance
Will Nicholas

Good cop, bad cop: Labor's fine line on AI penalties

The government plans to introduce an Australian regulator as part of its AI safety crackdown. (Joanna Kordina/AAP PHOTOS)

Australia wants artificial intelligence to behave, but might need it more.

Frontier AI companies could face criminal charges for failing to prevent harmful conduct, Prime Minister Anthony Albanese announced on Tuesday, as part of reforms triggered by OpenAI breaching a Medicare website.

Labor is yet to elaborate on specific penalties, but national security and AI expert Olivia Shen told AAP another piece of landmark legislation being debated in parliament might point the way.

Anthony Albanese
Prime Minister Anthony Albanese says AI firms may face criminal charges for failing to prevent harm. (Susie Dodds/AAP PHOTOS)

"We've actually mooted some pretty significant penalties under the digital duty of care," she said.

"I do think that the government is trying to use the Australian approach to the regulation of big tech in terms of social media ... as a bit of a benchmark."

The digital duty of care will impose fines of up to $109.2 million for social media companies who fail to protect users from harmful content on their platforms.

It aligns with beefed-up enforcement measures for the ban on under-16s from social media, which entered force in December 2025 but tightened in September.

"If you set it less than that, then you cop flack for giving AI companies a slap on the wrist," Ms Shen said.

"If you go over the top, it's like, 'hang on a second, that might be setting penalties at a time when we are still trying to attract AI investment'."

Labor will flesh out its AI safety crackdown, which includes a new national regulator, as part of its yet-to-be-legislated Australian AI Standards.

Much will depend on what the threshold for being a "frontier" AI firm was and how egregious the breaches were.

Mr Albanese did not say whether OpenAI's Medicare hack, which involved an unsupervised agent and did not compromise personal data, or sluggish disclosure would result in penalties under the emerging laws. 

If they only applied to deep-pocketed juggernauts like Google, Anthropic, OpenAI, XAI and their peers, a $109.2 million bill might just be a fly in the ointment.

Anthropic
Analysts say the government is keen to attract investment from AI companies such as Anthropic. (Dean Lewins/AAP PHOTOS)

Tech giants have coughed up sums in the order of $2 billion for breaching data protection regulations in the European Union, and taken the hit on the chin.

On the other hand, a smaller market like Australia was easier for AI firms to desert if they decided it wasn't worth the risk, Ms Shen said.

That EU regulation carried fines of up to four per cent of companies' global revenue for severe violations, but Labor has baulked at imposing similar penalties on social media companies in Australia.

They are also betting the house on AI to lift chronically stagnating productivity and put a lid on high inflation, meaning Canberra has more to lose from scaring tech giants away.

License this article

Sign up to read this article for free
Choose between a free or paid subscription to AAP News
Start reading
Already a member? Sign in here
Top stories on AAP right now