The federal government has released an exposure draft of its long-promised digital duty of care.
Author
- Rob Nicholls
Senior Research Associate in Media and Communications, University of Sydney
The headline feature is an "off switch". Social media platforms would have to ask users, through prompts that appear from time to time, whether they want to keep a personalised, algorithm-driven feed, or instead see mostly the accounts they have chosen to follow. The Prime Minister calls it "My feed, my way".
Communications Minister Anika Wells has been careful to say plenty of Australians appreciate the algorithm and will keep it. She argues big tech should offer them the choice in the first place, and should respect it.
Opposition Leader Angus Taylor says the Coalition has not seen the bill and wants parents to have more control over what their children see, but he has warned the government against censoring speech online.
But Shadow Treasurer Tim Wilson was more blunt, accusing the government of trying to "censor and silence Australians".
One Nation Leader Pauline Hanson is opposing the plan , saying "Australia should be moving away from government censorship, not building the Orwellian machinery for it".
Nationals Leader Matt Canavan also has reservations, telling ABC News Radio:
I'm deeply sceptical of giving any government, whether I was in charge, or Anika Wells is in charge, I'm deeply sceptical of giving any government the power to decide what you read on your phone.
But is that really what's being proposed in this policy?
Why do we need laws for this?
If algorithms cause harm, why do we need a statute? The obvious answer is negligence.
Someone who owes you a duty of care and fails to take reasonable care is liable for the damage that follows. This is a basic common law principle dating back to the early 20th century.
The problem is that negligence was built for disputes between two identifiable parties about one identifiable injury.
Algorithmic harm does not work like that. It is spread thinly across millions of people. It is a matter of probability rather than certainty because no single recommendation causes the damage.
Australian courts are also reluctant to open the door to claims where the number of potential plaintiffs is impossible to determine, particularly where the loss is purely financial. A claim against a platform would very likely fail on that ground before it got anywhere near the merits.
A statutory duty sidesteps the need for individuals to take cases to court. It's owed to users generally, it looks forward rather than back, and it's enforced by a regulator instead of by individual litigants.
Is an 'off' switch censorship?
The off switch is not censorship in any ordinary sense of the word. Turning off recommendations does not delete a post, suspend an account or stop anyone publishing.
Everything lawful stays lawful, and everything you follow still arrives.
The larger concern is about reach. The effect of changing how content is distributed changes who gets heard the most.
That is fair, but it cuts both ways. Your feed is already curated by a private ranking system you did not choose and cannot inspect. Giving the user that decision looks closer to the opposite of censorship.
Censorship is an authority deciding what may be said. This is an individual deciding what they will be shown.
A global fight about speech
The censorship framing lands at a moment when the platforms themselves have swung from harm reduction towards free speech absolutism.
Meta ended third-party fact checking in the United States in January 2025 and loosened parts of its hateful conduct rules.
X cut its trust and safety capability after 2022 and went to court against the eSafety Commissioner over a removal notice .
YouTube wound back election integrity policies and restored accounts it had previously banned.
American politics has amplified this. A House Judiciary Committee report accused eSafety Commissioner Julie Inman Grant of censoring Americans. Republican members of Congress wrote to the US Trade Representative on September 1 seeking a trade investigation into Australia's News Bargaining Incentive : a policy that would see tech companies pay for Australian news.
Free speech is now both a corporate defence and a trade lever. That is why an Australian safety law is being fought, particularly by the large platforms, as a speech law.
What censorship actually looks like
Real internet censorship has a recognisable shape. The state defines forbidden content, filters or controls the infrastructure, orders deletions on short deadlines, requires accounts to be tied to a government identity document , and punishes the speaker with little chance of review. China , Iran and Russia are the working examples.
The duty of care does not match that pattern. It creates no new speech offence. It does not punish users. It regulates systems rather than individual posts.
Two features still deserve close scrutiny.
It is important that businesses know what the law and the regulator expect. This helps ensure that large penalties do not give platforms a reason to remove too much, which produces censorship by private proxy rather than by government order.
It is also important that there is no scope creep. The legislation must be based on clear principles. For example, by defining the categories of harm, such as harm to children. Once that is done, the categories can't be changed by the regulator.
If the minister adds a category, it is a "disallowable instrument". This means the Senate can review the minister's decision and vote to overturn it. The regulator enforces against harms that the parliament has agreed.
Those are the parts to argue about. The off switch is not one of them.
Rob Nicholls is a member of the University of Sydney Centre for AI, Trust, and Governance. He receives funding from the Australian Research Council.
