Industry Minister Ed Husic flags crackdown on ‘risky’ artificial intelligence

A sweeping Artificial Intelligence Act will be considered by the Federal Government as it moves to rein in the technology’s potentially dangerous uses before they get out of control.

Industry and Science Minister Ed Husic will on Thursday announce plans to regulate tech companies designing and deploying AI in “high-risk” ways, such as facial recognition.

Creating a standalone AI Act — modelled on laws in Canada and the EU — is one of three options to implement the “mandatory guardrails” proposed in a new report released for consultation.

From today, new voluntary standards will also be able to help organisations safely use AI technology after new index revealed almost 70 per cent of businesses were failing to do so.

In an interview with The Nightly, Mr Husic said AI was an important part of everyday life with uses ranging from speeding up medical manufacturing to protecting customers from credit card fraud.

But Mr Husic said emerging forms of generative AI posed a risk that warranted intervention.

The report cited examples where AI software used to review resumes discriminated against women applying for software engineering jobs.

In another case, AI used to predict criminal behaviour undermined the presumption of innocence and showed bias against minorities.

“If you’ve got a big healthcare company like CSL that wants to use AI to accelerate drug development, that’s great,” Mr Husic said.

“If you’ve got another firm that’s using it to hire and fire people – that’s an issue.”

The report proposed a list of 10 “mandatory guardrails” that would apply to companies designing or deploying AI, which included a requirement to test AI models and monitor the system, and allow human oversight.

Countries around the world are rushing to regulate the nascent technology, with new UK Prime Minister Keir Starmer in July announcing plans for an AI Bill.

The Federal Government will consult for four weeks on the “mandatory guardrails” before finalising its position.

Mr Husic said he wanted to “move quickly’ to set up the new regime but wouldn’t be drawn on whether it would happen before the election, due in May.

“I wouldn’t want to make predictions on when we land this,” he said.

“We do want to send the signal to the public that we’ve listened, we’ve heard, we’re acting.”

The release of the new voluntary standards on Thursday comes as the latest Responsible AI index showed just 29 per cent of businesses were using AI safely and responsibly.

The Government intends to update the standards to respond to changes in best practices, matching the approach used in the EU, Japan, Singapore and the US.