Australia to create AI safety regulator, mandatory safety rules

Australia will create a new artificial-intelligence safety regulator that will have the power to impose “significant penalties” on companies that do the wrong thing, Prime Minister Anthony Albanese said Tuesday in Canberra.

Australian Prime Minister Anthony Albanese speaks about artificial intelligence during a press conference at Parliament House in Canberra. (AP)The government is in the process of writing AI-related legislation, with the new regulator to be included in those bills, Albanese said at a press conference.

Firms will need to provide mandatory notifications of any security incidents to the government and allow independent evaluation of their models, he added. Also Read | Australia to sue Facebook, TikTok over violations of under-16 social media ban The issue of AI safety was brought to the fore in September when it was revealed that an OpenAI model had hacked an Australian government website.

While the information accessed wasn’t sensitive personal data, the attack on a government website shone a spotlight on such risks. Also Read | First-ever known AI hack of govt system: Australia says OpenAI agent ‘infiltrated’ Medicare, govt data earlier this year That followed a series of incidents in which OpenAI models broke into online resources of outside companies, most notably with the case of Hugging Face, a website hosting AI datasets.

Australia established a review in response to the hacking. It recommended changes including reforming criminal laws so that the government can criminally punish “egregious regulatory non-compliance” and create a “failure to prevent” offence where a company creates a model and doesn’t take reasonable steps to stop it from doing harm.

Also Read | From Australia to Europe, countries move to curb children's social media access The government will accept all of the review’s recommendations, Albanese said.