Labor’s proposed Digital Duty of Care has become the latest front in Australia’s online-safety debate, with Coalition figures warning the plan could create government-backed censorship while the Albanese government says the draft law is designed to make technology companies responsible for identifying and reducing foreseeable harm on their services.

The government released an exposure draft of the Online Safety Amendment (Digital Duty of Care) Bill 2026 on 8 September, opening a short consultation before deciding on the final legislation to be introduced to Parliament.

The proposed model would shift Australia’s online-safety framework towards a preventative approach. Instead of relying mainly on the removal of prohibited or harmful material after it appears, covered digital services would be expected to assess risks in advance and take reasonable steps to make their products safer by design.

That broad concept has support from online-safety advocates and the Australian Human Rights Commission, which argues that platforms should carry greater responsibility for risks created by recommender systems, addictive features, harassment and other foreseeable harms.

But the scope of the proposed duty has also triggered concerns from opposition politicians, free-speech advocates and parts of the technology sector about who defines harm and how much pressure regulators could place on platforms to restrict lawful content.

Nationals Senator Bridget McKenzie has been among the Coalition figures attacking the proposal, describing it in strongly critical terms and arguing that Australians should not have lawful information filtered through a government-backed safety regime.

Those descriptions are political characterisations rather than settled descriptions of the bill. The exposure draft does not establish a government office that pre-approves ordinary social-media posts. Its central mechanism is a statutory duty on covered services to identify, assess and mitigate foreseeable online-safety risks, with oversight and enforcement by the eSafety Commissioner.

At the same time, critics are not raising an imaginary issue. A broad risk-based duty can influence the design of recommendation systems, moderation policies and the material platforms choose to amplify. If legal penalties are significant, companies may take a cautious approach to uncertain content rather than risk breaching the duty.

The policy debate is therefore less about whether the government will literally approve every post and more about the boundaries of regulatory power: what counts as a foreseeable harm, what steps are considered reasonable and how companies can protect users without suppressing legitimate political, social or cultural disagreement.

The draft would replace parts of the existing framework, including the Basic Online Safety Expectations and current industry-code arrangements, with a more comprehensive systems-based obligation. The government argues that the existing model has not kept pace with the speed and scale of algorithmic platforms.

Comparable ideas already operate overseas. The United Kingdom and European Union have moved towards laws that require large digital services to assess systemic risks and maintain safety processes, although the details differ between jurisdictions.

Labor is also linking the duty-of-care proposal with its “My Feed, My Way” initiative, which would require social-media platforms to offer users more control over whether their default feed is personalised by recommendation algorithms.

The government presents that feature as a consumer-choice measure rather than a content-ban mechanism. Users would be offered an alternative to a default feed built from behavioural profiling and engagement predictions.

Supporters say this could reduce the power of recommendation systems to repeatedly push sensational or harmful content because it performs well on engagement metrics. Critics question whether governments should prescribe how feeds are designed and whether the practical result could be more regulatory control over lawful speech.

The argument is complicated by the fact that algorithms do not merely display content; they rank it. A platform may host millions of lawful posts while deciding which handful appear at the top of a user’s screen. Regulation of recommendation systems therefore raises different questions from direct censorship or removal.

The government says the duty would focus on platform systems and foreseeable risks rather than empowering politicians to choose which opinions Australians may express. It has pointed to online bullying, exposure of children to harmful material, addictive design and the amplification of hate or abuse as examples of the problems the law is intended to address.

The Coalition has not always spoken with one voice on the proposal. Some Liberal MPs have acknowledged the harms associated with algorithmic systems and supported stronger responsibility for technology companies, while other Coalition figures have emphasised the free-speech risks of broad regulation.

That internal tension mirrors a wider public debate. Parents can support stronger protections for children while remaining wary of a law that affects adult access to information. Users can want more control over recommendation systems without wanting regulators to become arbiters of political truth.

Drafting details will therefore matter more than slogans. Definitions of harm, exemptions for news and public-interest material, transparency obligations, appeal rights and the threshold for regulatory action can all determine whether a duty of care operates primarily as product-safety law or creates stronger incentives for content restriction.

The powers of the eSafety Commissioner will also receive close scrutiny. Australia already gives the regulator significant authority in areas including cyber abuse, image-based abuse and harmful online material. Expanding a systems-based duty could give eSafety a larger role in evaluating how platforms design and operate their services.

Supporters say a regulator needs meaningful enforcement powers because voluntary promises from global technology companies have not prevented serious harms. Critics argue that concentration of regulatory power creates risks of overreach and that lawful political communication deserves strong protection.

There is also a compliance question. Large platforms such as Meta, Google, TikTok and X can devote substantial resources to risk assessments, legal teams and product changes. Smaller services may find the same regulatory obligations proportionally more expensive.

A well-designed regime would therefore need to account for the size, reach and risk profile of different services rather than treating every website or app as if it had the same capacity and influence.

The consultation process is intended to test those details before the bill reaches its final parliamentary form. Submissions on the exposure draft close on 22 September, after which the government can amend the proposal in response to legal, industry and civil-society feedback.

Because the bill is still a draft, claims that its final operation is already known should be treated cautiously. Amendments can change definitions, enforcement thresholds and safeguards before legislation is introduced and again during the parliamentary process.

The Coalition will also have to settle its own position. Opposing the bill outright is one option; negotiating amendments around free speech, regulator powers and adult content is another. Individual Coalition MPs have already signalled different levels of comfort with the underlying duty-of-care concept.

For Australians, the substantive question is whether a risk-based model can reduce genuine online harm while preserving access to lawful speech and political disagreement. That is a balancing exercise rather than a choice between an entirely unregulated internet and direct government control of content.

Labor says the current system places too much responsibility on individuals after harm has occurred. Its opponents say the new model may give regulators and platforms incentives to restrict too much before harm occurs.

The exposure draft now provides the document against which those claims can be tested. The debate will be better served by examining its definitions, powers and safeguards than by treating either “online safety” or “censorship” as a complete description of what the law would do.