Kate Chaney (Curtin, Independent) | Link to this | Hansard source

Think about what the internet could have been: a place where children learn, create and connect safely, where people in crisis find support and where algorithms surface what's true and valuable. Instead, children are being exposed to violent, sexual and misogynistic content. They're forming unhealthy emotional attachments to AI companions designed to maximise engagement. Adults are being radicalised by recommendation systems engineered to serve increasingly extreme content. This is the predictable consequence of design choices made by platforms that seek to maximise engagement and profit.

I commend the government for committing to a digital duty of care. This will require digital platforms, including social media sites and AI chatbots, to be safe by design. Digital platforms will have a legal responsibility to prevent harm, but the design of this commitment matters enormously, and I want to flag three things that the government must get right. First, there must be a single overarching duty. This would mean platforms had a general obligation to take reasonable steps to prevent foreseeable harm to users, rather than a list of named specific obligations. The difference is significant. A list of specific duties only captures what the drafters anticipated, not new and emerging harms. Specific duties for the highest risk categories can sit underneath an overarching obligation to give industry and regulators clarity, but the overarching duty must not be traded away.

Second, the framework must capture the full range of harms. The government's current issues paper sets the threshold at serious harm. But serious harm means loss of life or severe physical injury, and that threshold risks leaving out harms that are less acute or immediate but still deeply damaging and widespread. This could include the erosion of a child's developmental capacity through unhealthy relationships with AI companions, with a slow degradation of mental health and a teenager who spends years in an algorithmic environment that continues to serve up distressing content. None of these may clearly meet a serious harm threshold, but all of them are deeply problematic, and the government must lower the bar to harm, not serious harm.

Third, the eSafety Commissioner must be properly resourced and empowered. A well-designed duty of care on paper means nothing without a regulator capable of enforcing it. The commissioner needs powers to compel the production of documents and data from platforms to audit recommendation algorithms and to litigate appeals. It's great to see that the government has acted on this, passing laws this week to strengthen the commissioner's powers, but I remain concerned about the commissioner's budget and her ability to take on the tech giants. This will also be essential to an effective digital duty of care. Get these three things right and a digital duty of care can be genuinely transformative.