Safety is the label.
Control is the mechanism.
The Digital Duty of Care Bill gives Canberra and the eSafety Commissioner sweeping new power over what Australians can see, share and say online. It comes with penalties so large that platforms will err towards silencing contentious voices. Independent media gets less protection than the legacy press. We read all 91 pages. Here’s what it actually says.
Online Safety Amendment (Digital Duty of Care) Bill 2026
- Introduced
- 15 October 2026
- House
- Passed 28 October 2026
- Senate
- Committee report due 20 November 2026
- Regulator
- eSafety Commissioner
- Max penalty
- 60,000 penalty units per breach
Vote No
Maximum penalty for a company that breaches the “duty”, as stated by the Government
Increase in removal-notice penalties, from 500 to 6,000 penalty units
Down from 48 hours: how long a service gets to act on a complaint before eSafety can issue a removal notice
Length of public consultation on the exposure draft
10 November 2026 · 8 minute read
Nobody wants children harmed online. That’s exactly why bills like this one pass. The Albanese Government’s Online Safety Amendment (Digital Duty of Care) Bill 2026 is sold as a child-safety measure. Read past the press release and into the Bill itself, and you find something much bigger. It’s a permanent system of state-supervised speech management. It’s run by an unelected Commissioner, it can be expanded by ministerial decree, and it’s enforced through penalties big enough to make every platform in the country flinch.
The Bill passed the House of Representatives on 28 October 2026. It is now with the Senate, and the Senate is where it can still be stopped.
A "duty" with no edges
At the centre of the Bill is a single sentence. A person responsible for an online service “must take appropriate action to ensure a safe online environment” s 26(1). Breaching that duty carries a civil penalty of 60,000 penalty units s 26C. At today’s value of $364 a unit, that’s $21.84 million for an individual. The Government’s own website says it’s up to $109.2 million for a company.
Who is caught? Far more than Facebook and TikTok. The Bill’s list of “online services” covers social media, messaging, search engines, app stores, hosting services and AI tools. It also covers “designated internet services”, the category that ordinary websites fall into. It even reaches businesses that make or install internet equipment s 25B(1). Small and low-risk services escape only if the Minister decides to exempt them by legislative instrument s 25B(3). Your protection depends on ministerial grace, not on the law.
One line deserves special attention. When deciding whether a service has done enough, “the cost of the action or measures must not be taken into account” s 26A(3). A garage forum, a community website and a trillion-dollar tech company all face the same open-ended obligation, and none of them can argue that the burden is unreasonable.
Who decides what's "harmful"?
The Bill lists “seriously harmful” material for all Australians s 25D. Most of that list is already illegal, like child abuse material, terrorism and death threats. Nobody is defending any of it. The danger lies in the second list, material “harmful to children” s 25E. Every platform must protect children from it, and it includes:
“material or conduct that encourages or promotes hostile attitudes towards women or gender equality” s 25E(1)(c)
What counts as “hostile” to “gender equality”? Is it opposition to gender quotas? Questioning gender-identity policy in schools? Criticising a particular equality program? The Bill doesn’t say. That judgement falls first to platform moderators facing nine-figure penalties, and then to the eSafety Commissioner. The list then ends with a catch-all: “any other material or conduct… that may cause serious harm to children” s 25E(1)(f). The Explanatory Memorandum says outright that the list is not meant to be exhaustive.
Platforms can’t reliably tell an adult from a 15-year-old. So the practical way to “protect children” from a category of speech is to restrict or bury it for everyone. Parliament won’t be banning those views. It will be making them commercially radioactive.
The list is built to grow. The Commissioner may advise the Minister on “emerging categories of harm” on the Commissioner’s own initiative s 28C. The Australian Human Rights Commission has already asked the Senate to consider whether hate speech should be added to these lists at the first review. Today’s limits are just the starting point.
Censorship by proxy
The Government says the Bill regulates “systems and processes”, not individual posts. That’s the trick. Canberra doesn’t have to take down your post. It only has to make a platform liable for tens of millions of dollars if its “risk assessment” doesn’t satisfy the Commissioner. The platform’s lawyers will do the rest. Every rational compliance team responds the same way. It removes more, ranks contentious accounts lower, and stops recommending anything that could be called “hostile”.
The Bill gives the Commissioner the tools to steer that process directly. The Commissioner can set the “standards, benchmarks or other metrics” platforms must use in their risk assessments s 27(2)(f), 27(4). The Commissioner can compel information about the algorithms that “select, order, recommend, promote, restrict or moderate material, accounts or search results” s 28A(3). The Commissioner can issue “remedial directions” that are expressly not legislative instruments, which means Parliament can’t disallow them s 26E(4). And the Commissioner can publish “guidelines” on what the duty requires s 28.
So a government regulator gets a hand on the dial of what Australians are shown in their feeds, and nobody in Parliament votes on where it gets turned.
Two classes of speech: theirs and yours
The Bill exempts “professional news material” s 25F(1)(d). That sounds reassuring until you read the definition. To qualify, a publisher must be bound by the Australian Press Council or Independent Media Council codes, the broadcasting codes, the ABC or SBS codes, or “analogous” internal standards relating to “quality journalism”. It must also have “editorial independence from the subjects of the person’s news coverage” s 25F(3).
That’s a protected class built around the legacy media. Independent outlets, citizen journalists, podcasters, commentators and advocacy organisations like TPAUS are not automatically inside it. We hold governments to account and speak for Australians who don’t see themselves in the mainstream press. Our best hope is a general carve-out for material “in the public interest” s 25F(1)(i), and the Bill doesn’t define that phrase. Platforms will interpret it inside their risk models before any court ever does.
The result is a two-tier system. Credentialed outlets can say it, and everyone else has to hope an algorithm agrees it was in the public interest. Independent voices that challenge government narratives are exactly the ones who lose that bet. Nobody has to target them by name. The incentives do the targeting.
Sock puppets, secret examinations and shared files
Some parts of the Bill have nothing to do with platform design and everything to do with state power.
- Government fake accounts. The Commissioner may operate “sock puppet” accounts, defined as “a false or fictitious identity”, on any regulated service, despite the platform’s own terms s 205G, 205J. They can be used to observe, download and record material and to engage with services. Anyone acting in good faith is immune from civil proceedings s 205L.
- Compelled examination in private. If the Commissioner believes any person can give information relevant to “a possible contravention of this Act”, the Commissioner can summon them and examine them on oath. The examination takes place in private s 198–201. Refusing to answer carries up to 12 months’ imprisonment s 204, and the privilege against self-incrimination is removed s 205. Journalists get a narrow exception that protects confidential sources and nothing more.
- Your information, shared widely. The Commissioner’s power to disclose information is widened to any Commonwealth entity s 212, as amended.
Law made outside Parliament
On the core questions, the Bill hands lawmaking to the executive. The Minister can decide by instrument which services are exempt s 25B(3) and which count as “social media” s 25C(2). The Minister can declare new design features to have “negative behavioural impacts” s 25G(2) and force platforms to build specified tools s 27A. The Commissioner sets the risk-assessment rules and complaint-process rules s 27(4), 27G(3). The architecture is designed to change without coming back to Parliament for a vote.
And the process has been rushed. The exposure draft was open for comment for 14 days. The Bill was introduced on 15 October and passed the House 13 days later. The Senate committee has until 20 November to report. The Government wants it law by Christmas.
We've seen this before
In 2024 the Government was forced to abandon its misinformation bill after it failed to win Senate support. This Bill takes the same goal of a regulator deciding which online content is acceptable and achieves it by a different route. Instead of labelling speech “misinformation”, it labels it “harmful”. Instead of fining users, it fines the platforms that carry them. The outcome for anyone outside the approved consensus is much the same.
Children deserve protection, and much of what this Bill targets is already a crime. Parents deserve better tools. None of that requires giving a single unelected official, backed by nine-figure penalties and ministerial decree, effective control over the visibility of political speech in this country.
The Senate can still stop this. Tell your Senators to vote no.
The Bill, clause by clause
Every claim on this page comes from the Bill as passed by the House on 28 October 2026. Check it yourself.
s 26 & 26C
The open-ended duty
Every online service must “ensure a safe online environment”. A breach carries a penalty of up to 60,000 penalty units.
s 26A(3)
Cost is irrelevant
The cost of compliance “must not be taken into account” when judging whether a service has done enough.
s 25E(1)(c), (f)
Vague "harm" categories
These include “hostile attitudes towards women or gender equality” and a catch-all for anything that “may cause serious harm”.
s 25F(3)
Two-tier news exemption
The exemption protects “professional news material” bound by Press Council or broadcast codes. Independent media isn’t automatically covered.
s 27(4), 28A
A hand on the algorithm
eSafety sets risk-assessment metrics and can demand details of how feeds rank, restrict and moderate.
s 205G–205L
State sock puppets
The Commissioner can run fake identities on platforms, overriding the platforms’ terms, with immunity from civil suit.
s 199–205
Secret, compelled examinations
Private examinations on oath. Refusal carries up to 12 months’ jail. No privilege against self-incrimination.
s 25B(3), 25G(2), 27A
Rule by instrument
The Minister can expand who and what is covered without a new Act of Parliament.
Sch 1, Pt 4
Removal penalties up 12x
Removal-notice penalties go from 500 to 6,000 penalty units, and the complaint window before a notice shrinks from 48 to 24 hours.
Be ready when the Senate votes.
The Bill has passed the House and is now before the Senate. The vote could come with very little warning.
Enter your name, postcode and email. We’ll email you as soon as the vote is scheduled and tell you exactly who to contact and what to say, while it can still make a difference.
Get the vote alert
Questions people ask
Parts of it are, and much of the material it targets, such as child abuse material and terrorist content, is already illegal under existing law. The problem is everything else it bundles in: vague categories of “harm”, a regulator-controlled risk system, government fake accounts, compelled secret examinations and power to expand the regime by ministerial instrument. Children can be protected without building that machinery.
Not in so many words, and the Government will say so. But it doesn’t need to. It makes platforms liable for up to $109.2 million if the regulator decides their “risk assessment” falls short, and it leaves terms like “hostile attitudes towards… gender equality” undefined. The predictable result is that platforms over-remove and quietly bury contentious political content to stay safe.
“Professional news material” is exempt, but only from publishers bound by the Press Council, broadcasting or ABC/SBS codes or “analogous” standards. Independent and citizen media are not automatically inside that definition. They have to rely on an undefined “public interest” exemption that platforms will apply first.
The new removal and disclosure powers in Schedule 1 start the day after Royal Assent. The duty of care itself starts 12 months after Royal Assent, when the existing industry codes and Basic Online Safety Expectations are repealed.
Sign up for the vote alert above. The Bill has already passed the House, so the Senate is the only place left to stop it. We’ll email you when the vote is scheduled, with exactly who to contact and what to say.
Sources
- Online Safety Amendment (Digital Duty of Care) Bill 2026, as read a third time, House of Representatives, 28 October 2026.
- Explanatory Memorandum, Online Safety Amendment (Digital Duty of Care) Bill 2026, circulated by the Minister for Communications.
- Department of Infrastructure, Transport, Regional Development, Communications, Sport and the Arts, Digital Duty of Care (maximum penalty figure).
- Commonwealth penalty unit value of $364 from 1 July 2026, Crimes Act 1914 s 4AA.
- Australian Human Rights Commission, Submission to the Environment and Communications Legislation Committee, 30 October 2026.
