Can AI be held legally responsible for its actions?

Australia has introduced AI safety standards and policy proposals, but legal accountability remains a grey area. Here’s where things stand in 2025.

Transcript

00:00Can artificial intelligence be held legally accountable if it causes harm,

00:05spreads misinformation, or makes a flawed decision? It's a question lawmakers around

00:10the world are scrambling to answer. Under Australian law, AI is not a legal entity,

00:20which means it cannot be sued, fined, or imprisoned like a person or a corporation can.

00:25In 2024, the European Union introduced the world's first comprehensive AI act,

00:32which includes strict rules around accountability and transparency.

00:35It's now pretty much smarter than humans. And that's the dangerous part. There are no rules,

00:45no laws, no regulations that govern AI. It is a wild west. And who is leading it?

00:53It's the tech giants. And have the tech giants demonstrated to date that they are ethically

01:00driven and purposeful in their mission? No, they haven't. In early 2025, the federal government

01:06released a discussion paper proposing tougher regulations around high-risk AI systems and the

01:12introduction of a new AI safety commissioner. From September 2024, Aussie government agencies were

01:20also required to publish transparency statements about their AI use and appoint officials responsible

01:26for its safe deployment. The research that we're releasing today from the National AI Centre

01:32shows that nearly 80% of businesses in Australia think they're doing the right thing, but only around 30%

01:39are putting in place the responsible practices required to use AI. So what we need to do

01:44is create that bridge between best intention and best practice. Currently, most laws still treat

01:51AI as a tool, with responsibility falling on the humans or organisations that use or develop them.

01:57For example, if a self-driving car crashes, the company behind the software may be liable,

02:02and not the AI itself. We have to be very careful that what we tell the AI we want is what we actually

02:10want because what AI does is ruthlessly pursue those goals that we give it. And if those goals are even

02:17slightly misaligned with our own, we could end up with some really problematic consequences.