New paper proposes mandatory ‘guardrails’ to test AI models

Mandatory rules for artificial intelligence are part of a planned government crackdown to combat the misuse of high-risk technology.

Transcript

00:00The Federal Government says that the use and misuse of AI technology is one of the most

00:08complex issues facing lawmakers around the world.

00:12And it's already very pervasive throughout our very busy lives.

00:16For example, we use chat bots, facial recognition, those assisted software tools, and on the

00:22darker side, the deepfake images that we see.

00:26The Federal Government believes that by introducing 10 mandatory guardrails, including the prohibition

00:32of some of the most high-risk AI tools, that can actually help prevent some of the damage

00:38and harm caused by its misuse.

00:40Research that we're releasing today from the National AI Centre shows that nearly 80% of

00:45businesses in Australia think they're doing the right thing, but only around 30% are putting

00:51in place the responsible practices required to use AI.

00:55So what we need to do is create that bridge between best intention and best practice.

01:01That's what we're trying to do with the work we're releasing today around the guardrails,

01:05which is identify the risks and make business aware of those risks, and then show them the

01:10guardrails, the ways in which they can deal with those risks.

01:14And Federal Industry Minister Ed Husic there is referring to the Responsible AI Index.

01:20This was a survey that was sponsored by the Government.

01:22Now he says that while there is this four-week consultation period, which will be seeking

01:28information and feedback from industry, as well as working out what sort of penalties

01:32might apply, there is right now something that is happening with immediate effect, and

01:37that is a voluntary AI safety code and the ability for people to be able to go and seek

01:42help if they feel that they have been damaged by the use of AI.