The coalition has launched a petition to kill the bill. Apparently, the Online Safety Amendment (Digital Duty of Care) Bill 2026 is government censorship. That is an extraordinary description of legislation that gives Australians more control over what they see online while requiring technology companies to take responsibility for the products they design. I've been on Facebook since around 2008 and on Myspace before that. I'm starting to give away my age! Then again, back then it was a terrific way to keep in touch with friends and family, particularly while I was travelling overseas. When I returned to Australia, it helped me stay connected with the people I'd met along the way. I could see their photographs, hear their news and maintain friendships across thousands of kilometres. It was simple: you followed someone because you wanted to hear from them, and, for the most part, you did.

These days, Facebook is a different experience. Posts from friends and family are still there, but they're sprinkled throughout my feed like salt and pepper in between a sponsored post, videos from strangers, arguments I never asked to join and advertisements for things I have no intention of buying. I joined Facebook to keep up with people I actually know; now, I spend more time scrolling past people I don't. Apparently the algorithm knows what I want better than I do.

The technology writer Cory Doctorow has explained how platforms arrive at this point. At first, a service offers something genuinely useful. It connects friends, helps small businesses find customers and allows people to share information. We join because it works. Then, when enough people depend on it, the company starts prioritising advertisers and paying customers. The experience changes. What we choose to see competes with whatever keeps us looking at the screen. Eventually, the platform starts squeezing those businesses too. A local business spends years building an audience only to discover it may have to pay to reach the people who already chose to follow it. The experience gets worse, but leaving becomes harder because our friends, customers and communities are all still there. That is not just a complaint about the advertisements; it is about the enormous power platforms acquire over connections Australians have built themselves.

It brings us to My Feed, My Way. Under the proposed rules, Australians over 16 will be able to choose whether their primary social media feed contains posts from accounts they follow or personalised recommendations selected by an algorithm. If you like the recommendations, you could opt in. If you prefer posts from friends, family and chosen creators, you could choose that instead. The choice should be clear, easy to make and persistent. A platform should not quietly reverse it each time you open the application or make the alternative impossible to use. It should not take a science degree, six settings screens and the patience of a saint to see what your friends are posting.

The detailed feed requirements are proposed through a legislative instrument enabled by the bill, not spelled out word for word in the primary legislation. That distinction's important and parliament can scrutinise how the rules operate, but the underlying principle is straightforward: the person holding the phone should have a meaningful say in what appears on it. The coalition calls this censorship, but I struggle to see how giving someone the choice to reject recommendations silences the person whose content is not recommended. Nobody has the right to appear in a stranger's feed simply because a company's algorithm predicts that an argument will generate engagement. And, if choosing your own feed is censorship, what should we call a system in which a technology company chooses it for you?

There is another question we need to ask. For years, social media companies have presented themselves as platforms where people post content, but those services are not passive noticeboards. They select, rank and recommend posts. They decide which video plays next, which controversy is amplified and which content reaches people who never asked to follow its creator. Those are decisions about distribution. They can have commercial and human consequences. Traditional publishers carry legal responsibilities associated with what they publish. The High Court's decision in Voller demonstrated that facilitating third-party comments can, in particular circumstances, amount to publication for defamation law. That decision did not declare every technology platform a publisher of everything it hosts, and this bill does not do so either—but the policy question remains. When a business activity recommends and amplifies content for profit, what responsibility should accompany that power? There is a difference between allowing someone to speak and deliberately putting their words in front of a million strangers.

I do not suggest that every recommendation amounts to an editorial endorsement—a newspaper editor and a recommendation system are not identical—but we cannot ignore the decisions made by the people who built and tune those systems. They determine the incentives, the measures of success and the way content travels. If a company can optimise its system to maximise advertising revenue, it can also be required to assess how that system exposes people to foreseeable harm. That is a more useful discussion than pretending the platform has no role once someone presses 'Publish'. The digital duty of care begins to address the risks created by the system doing the amplification, and the principle should be familiar to anyone who has worked in an industry where safety is taken seriously. In workplaces across Australia, employers have duties to identify hazards and control risks. They cannot simply wait for a worker to be injured, hand them a complaint form and declare the system successful. We apply similar expectations to cars, food and children's toys. If a product has a foreseeable defect, the company designing it has a responsibility to act. Online services should not get a different rule simply because their products apply on screens.

This bill requires providers to assess the risks arising from their services, including risky design features, and take effective measures to reduce them. It is about prevention, not just removing a harmful post after somebody has already been hurt. It also changes the way online safety is regulated. Today there are 17 industry developed codes and standards spread across different services and obligations. The bill replaces that complicated framework with an overarching duty of care after a transition period. Rather than waiting for the industry to draft a new code each time technology changes, providers must assess new risks and respond to them. Companies remain responsible for designing their services, but safety is not simply left to rules written by the companies being regulated. The duty applies to specified serious harms affecting adults, including child sexual exploitation, explicit threats of violence, intimate image abuse and material encouraging suicide or serious self-harm. Children receive additional protections against pornography, material encouraging eating disorders, dangerous stunts, bullying and hostility towards women, as well as design features associated with harmful behaviour. This is not about outlawing political disagreement; it is about defined categories of harm and appropriate steps to reduce the risk that online services expose Australians to.

The government has already introduced the social media minimum age of 16. According to the updated government figures, more than five million underage accounts have been deactivated, removed or restricted. That does not mean the job is finished. There are still underage accounts, and the eSafety Commissioner is investigating possible noncompliance by major platforms. (Quorum formed) The answer to noncompliance is to enforce the law, not to abandon the expectation that that platform should follow it. The figures also describe survey respondents reporting more sport and better sleep, alongside reduced exposure to some harmful experiences. These are early findings, not a declaration of victory. The responsibility remains on platforms to meet the law and on regulators to test what is happening in practice. The risks to children extend beyond social media accounts. They use online games, messaging services and artificial intelligence chatbots, including products that imitate a friend or adviser. Parents cannot expect every interaction or understand every feature of every new product. Children should not be expected to recognise unsafe design before the provider has any responsibility to intervene.

The bill also addresses a particularly disturbing use of artificial intelligence: apps that generate fake nude images without consent. A perfectly ordinary photograph can be turned into an intimate image the person never agreed to create. For the person targeted, especially a child, the damage can be devastating. The commissioner will have powers to require action against apps and websites predominantly designed and used for generating that material. When serious cyberbullying or adult cyberabuse is reported, the bill reduces the waiting period before the commissioner may intervene in relevant cases from 48 hours to 24 hours. Australians should not have to submit a complaint into a digital black hole and wonder whether anybody has read it.

There is another issue that those opposite have become particularly excited about: sock puppet accounts. It sounds sinister, but in reality these regulated test accounts are rather like mystery shoppers for the internet. If a shop says it checks identification before selling an age restricted product, an inspector can test whether that actually happens.So why should regulators have to accept a technology company's word that its safety systems work? Can an adult profile contact a child through an online game? Can a younger teenager open an account without age checks? How does an AI service respond to attempts to produce abusive images? Those are practical questions. A test account can help answer them.

These provisions allow appropriately regulated testing and provide a pathway for approved researchers to obtain relevant safety data. They are not a licence for government to plant political opinion in people's newsfeeds. They are a means of checking compliance. We do not let companies mark their own homework, and there is no reason to start with big tech.

The coalition has also criticised powers that were removed before this bill was introduced. The proposed ministerial power to add new categories of 'harmful material' was removed. The final bill contains protections for professional news material. It does not establish a general regime for deciding whether political claims are true or false. Those are facts about the legislation before parliament, whatever a petition might suggest. Nor does the duty mean a neighbourhood florist's basic website will be regulated as if it were one of the world's largest social media platforms. The framework is risk based, with minimal obligations intended for low-risk services and guidance to be provided by the eSafety Commissioner. Technology should serve the people who use it, not the other way around. I commend the bill to the House.