Australia’s proposed Digital Duty of Care laws are becoming a major political battleground, with the Greens arguing that social media companies should face stronger legal obligations over harmful content and algorithmic amplification while the Coalition warns the plan could create excessive government control over online speech.

The federal government released an exposure draft of the Online Safety Amendment (Digital Duty of Care) Bill 2026 on 8 September, opening public consultation until 22 September. The proposal would place a proactive duty on major digital platforms to identify and reduce risks to users rather than relying only on complaints after harm has occurred.

One of the most visible elements is a plan to give users greater control over recommendation algorithms. Communications Minister Anika Wells has said platforms would have to offer people a clear choice between an algorithmically selected primary feed and a feed based on accounts they have chosen to follow.

Supporters of the reform argue that recommendation systems can intensify harmful material by repeatedly serving users content that keeps them engaged, even when the material relates to bullying, eating disorders, misogyny, violent material or other risks. The government says the objective is to shift responsibility from individual users towards the companies designing and operating those systems.

Greens senator David Shoebridge has been among those calling for stronger platform accountability. In Senate debate on 8 September, he argued that “fixing the feed” and imposing a duty of care on platforms was a better path than relying on tougher penalties associated with the government’s existing social media minimum-age regime.

Shoebridge has also been appointed the Greens spokesperson for Digital Rights and AI. In announcing the role, the party said it wanted the duty of care framework to address emerging risks from AI chatbots and agents as well as conventional social media platforms.

That does not mean the Greens have offered unconditional support to every element of Labor’s draft. The party has pressed for stronger parliamentary oversight and safeguards around how harmful material is defined, while some Greens representatives have argued users should receive stronger default protections from algorithmic amplification.

The Coalition has taken a more sceptical position. Its representatives have raised concerns about censorship, broad definitions of harm and the possibility that ministers could gain too much influence over what platforms are required to restrict. Labor has responded by offering to reduce ministerial discretion and give Parliament a greater role in determining future categories of harmful content.

The dispute is important because claims that the proposed law is designed to suppress a particular political viewpoint go beyond what the draft itself establishes. The legislation is framed around platform systems, risk assessment, online safety and user controls. Politicians from different parties have nevertheless argued over whether the practical effect could advantage or disadvantage particular forms of political messaging.

Shoebridge has been sharply critical of One Nation and the Coalition in separate parliamentary debates, accusing the Coalition of following One Nation’s lead on some votes. Those political attacks provide context for partisan arguments around the duty-of-care debate, but they should not be treated as proof that the legislation itself is designed to stop conservative speech.

The exposure draft is broader than political content. Government material says platforms would be expected to manage risks associated with illegal content and, for children, additional categories of psychological and social harm. The model also gives users tools to change the way recommendation systems shape their feeds.

Prime Minister Anthony Albanese and Wells have described the algorithm choice as a consumer-control measure. Wells has said users who enjoy algorithmic recommendations could continue using them, while those who prefer to see posts from people and accounts they deliberately follow could switch to that model.

The government is also trying to learn from overseas regulation. European reforms have already required some large platforms to offer alternatives to personalised recommendation systems. Australian ministers say they want to avoid situations in which companies technically comply while making the non-algorithmic option difficult or unattractive to use.

Critics argue that imposing a legal duty of care on online services creates difficult questions about how companies judge harm. Online platforms host enormous amounts of lawful but controversial material, and rules designed to reduce serious risks can affect political discussion if definitions are too broad or enforcement incentives encourage companies to remove content defensively.

That concern is one reason the draft process matters. The government has not yet reached the final form of the bill, and the consultation period allows civil-liberties organisations, technology companies, safety advocates, academics and ordinary users to propose changes before legislation is introduced to Parliament.

Supporters of the bill respond that the status quo also involves significant private power. Recommendation algorithms already decide what users see, but those systems are designed by companies and are often optimised for engagement. Requiring greater transparency and user choice, they argue, can reduce corporate control rather than increase government control.

The debate therefore cannot be reduced to a simple choice between free speech and censorship. It involves competing questions about corporate responsibility, government power, personal choice, child safety and the architecture of online platforms.

For the Greens, the central argument is that the law should focus more directly on the systems that push content towards users. Shoebridge’s Senate comments emphasised platform accountability rather than relying only on age restrictions. Other Greens figures have similarly argued that people should have greater control over whether algorithms determine what appears in their feeds.

For the Coalition and some civil-liberties critics, the concern is that a broadly framed duty could become a mechanism for overreach if future governments or regulators expand definitions of harm. They want tighter limits and clearer safeguards before supporting the legislation.

Labor is attempting to bridge those positions by retaining a strong duty on platforms while adjusting oversight arrangements. Whether that produces enough support in the Senate remains uncertain. The Greens could provide a pathway if the government accepts amendments, while the Coalition is pressing for substantial changes.

The final law, if passed, will matter beyond the current political fight. It could change the design of social media feeds, require major platforms to conduct regular risk assessments and create substantial penalties for failures to meet safety obligations.

For now, the most accurate description is that Australia is considering a far-reaching online safety regime whose exact boundaries are still being negotiated. Claims that it will either “fix” social media or silence political opponents go further than the evidence allows. The exposure draft establishes a framework; Parliament, regulators, platforms and ultimately the courts would determine how that framework operates in practice.

The practical effect on users will depend heavily on implementation. A choice screen is only meaningful if the alternatives are easy to understand and if choosing a following-based feed does not remove unrelated features or repeatedly push users back towards algorithmic recommendations. Regulators would therefore need to examine not only whether a button exists, but whether platform design respects the choice in practice.

Enforcement will be another test. Large technology companies operate across many jurisdictions and can change products rapidly. A statutory duty may provide stronger leverage than voluntary codes, but regulators will still need technical expertise, access to information and penalties large enough to influence companies whose global revenues dwarf those of many Australian businesses.

The consultation period gives Parliament an opportunity to resolve those details before the bill is finalised. The strongest version of the reform would need to protect users from demonstrable harm without creating a vague political-content test. That balance, rather than partisan claims about whose messages are amplified, is likely to determine whether the legislation gains durable public support.