> ## Content Index > Fetch the complete content index at: https://www.theoceaniacables.com/llms.txt > Use this file to discover other available public pages before exploring further. # Australia Drafts Online Duty of Care; Critics Warn of Censorship - URL: https://www.theoceaniacables.com/australia-drafts-online-duty-of-care-critics-warn-of-censorship/ - Published: 2026-09-08T21:57:29.000Z - Updated: 2026-09-08T22:15:45.000Z - Description: Australia's online safety law would fine platforms $109.2m and let a minister decide what's "harmful" online, critics call it stealth censorship. - Author: The Oceania Cables - Tags: Australia, Legislation, Policy ### **Summary:** On 8 September 2026 the Albanese Government released an exposure draft — a draft law published for comment before it is introduced to Parliament — of the Online Safety Amendment (Digital Duty of Care) Bill. Targeted consultation is open now until 12pm on 22 September, with a bill promised later in the year. The draft would place a legal duty of care on a wide class of online services — not only social media, but messaging, games, dating, search engines, app stores, hosting and many generative-AI tools. It would replace the current non-binding safety expectations with a fineable obligation: to take reasonably practicable steps to keep users in Australia safe from listed harms and from product designs the statute treats as harmful. The launch led with a consumer feature, "My Feed, My Way," and the duty is backed by civil penalties running into nine figures. The draft's most consequential feature is structural. It would let the Communications Minister add new harm categories and new "harmful" design features by legislative instrument — a rule Parliament can disallow, but which requires no new primary legislation — provided the minister is satisfied the material *may* cause serious harm. The government says the bill is not censorship and is not a revival of the misinformation bill it withdrew in November 2024\. Opposition Leader Angus Taylor calls it a likely attempt to censor; Deputy Liberal Leader Jane Hume says the minister should not decide what counts as online harm; Nationals leader Matt Canavan says "we should never give censorship powers to people who are proven liars"; and One Nation's Pauline Hanson says the powers would push platforms to "over-censor." ### **Detailed Report** ### **1\. The Launch: Feed Choice Out Front, Nine-Figure Fines Behind** At a Canberra press conference on 8 September, Prime Minister Anthony Albanese and Communications Minister Anika Wells staged the launch around a feature, not a penalty. The centrepiece was "My Feed, My Way": the government said users aged 16 and over would be asked whether their default primary feed should be an algorithmic recommender or a chronological following feed, and could switch later. Wells called it a neutral path through the middle and rejected the Greens' model, under which the algorithm would stay off unless a user turned it on. The framing was consumer choice, not government control. "This is not about giving government control," Albanese said. "It's about giving people control. It's about putting choice back into the hands of Australians online." Wells reached for the language of consumer safety, comparing the regime to standards for cars, food and appliances: tech companies could still move fast, she said, "but they will no longer be allowed to break things." The enforcement sat behind the feature. Albanese said platforms that ignored the law would face "significant penalties" of more than $100 million; the government's written figure for a single breach of the duty is up to $109.2 million. Wells added that a company breaching both the duty and the existing under-16 social media ban would face fines that "ratchet up to more than $200 million per offence." The launch was staged through lived experience. Bereaved families joined the ministers on the platform, there to put faces to the harms the bill names — suicide and self-harm content, bullying, and the eating-disorder material it would require platforms to act on. The government also pointed to its record under the under-16 ban — more than five million underage accounts removed or deactivated — and cast Australia as the front of a global movement. ### **2\. What the Bill Does: A Civil-Penalty Duty Across Online Services** The exposure draft, dated 7 September, would replace the Basic Online Safety Expectations in the Online Safety Act 2021 with a civil-penalty duty to provide, so far as is reasonably practicable, a safe online environment. The expectations were a ministerial standard a platform could be asked to report against. The duty is something a platform can be fined for breaching. **Who would be covered** • Social media, messaging, games and dating services. • Search engines, app stores and hosting services. • Many generative-AI tools that create material and distribute it through those channels. • Some device and operating-system activity in Australia tied to those services. • The minister may exempt a service or class that poses little risk or is barely used here. **What a covered service would have to do** • Manage its design features, including any user-control tools the minister requires. • Write a risk assessment at least once a year, and again before any change that could add risk; keep it for six years; and produce it within 30 days if the eSafety Commissioner — the statutory online-safety regulator — asks. • Take effective measures against the risks the assessment identifies. • Failing any one of these limbs breaches the duty. **Harms named for children** • Pornography. • Instruction in disordered eating. • Material that "encourages or promotes hostile attitudes towards women or gender equality." • Content glorifying crime or dangerous stunts. • Abuse, harassment or bullying. • Anything else that could seriously harm a child — and anything else the minister adds. **Harms named for all users** • Child sexual exploitation, grooming and sexual violence. • Extreme cruelty; rape and death threats; menacing or harassing conduct. • Instruction in suicide, self-harm or animal cruelty. • Support for a listed terrorist organisation, or praise of a terrorist act. • Instruction in a criminal offence, in illicit drug use, or abhorrent violent conduct. • Any other material the minister determines may cause serious harm. **Product features the statute treats as harmful** • Recommender feeds, infinite scroll, likes and view counts, disappearing content and login-gating. • Any further feature the minister adds to the list. • On social media, these must not operate for under-16s in Australia; for adults they become the subject of the feed-choice tools and the risk assessments. Running alongside the duty is a separate, faster schedule: new notice powers against "nudify" apps and search engines, and a cut in some takedown clocks from 48 hours to 24\. The draft would also repeal the online-content industry codes now sitting under the 2021 Act, so future rules would no longer pass through a negotiated code cycle. ### **3\. The Minister's Power to Decide What Counts as Online Harm** The heart of the draft is a set of ministerial powers, each exercisable by legislative instrument — a rule Parliament can disallow, but which needs no new bill. By instrument, the minister could decide who is covered, what counts as social media for the under-16 design shutdown, what else is a serious harm for all users, what else is a child harm, which further product features are treated as harmful, and which user-control tools must be built. The threshold for adding an all-user harm is that the minister is *satisfied the material may cause serious harm* — not that it does. It is a low bar, and it is the government's own description of the test. Wells has already used this style of instrument once, to pull YouTube into the under-16 social media ban. The Senate could have disallowed that designation. Nobody moved to — and that matters, because disallowance is the only parliamentary check on the power, and on the one occasion it might have bitten, it did not. Asked whether Labor would trade the add-on power away to win Coalition support, Albanese would not. Wells argued the power was needed for speed, citing 29 December, when — she said — the AI service Grok began letting users generate nude images from photos, and existing law could not stop it. Under the new power, she said, she could have consulted the Commissioner, declared the nudifying of images of teenagers a new serious harm, and faced any Senate disallowance in February. Her question to the Coalition: would they prefer to "recall parliament every time there's a new Grok"? ### **4\. Labor's Internet Record: Misinformation, Age Bans and the Duty of Care** The duty does not arrive cold. It is the latest move by a government that has reached for online-content and platform powers repeatedly — winning some fights, losing the one that touched speech most directly — and each earlier step was sold as the measure needed to close the last remaining gap. The clearest precedent is the one that failed. Two Combatting Misinformation and Disinformation bills would have let the communications regulator approve enforceable industry codes on "seriously harmful" false content. They collapsed against a wall of Coalition, Greens and crossbench opposition; Elon Musk's one-word verdict on the revised 2024 text was "Fascists." Communications Minister Michelle Rowland withdrew the bill on 24 November 2024, conceding "no pathway," and Albanese called it "the end of it." It is that defeat the government is most anxious to distinguish now, stressing that the duty is a different legal instrument and that the words *misinformation* and *disinformation* appear nowhere in the text. Where misinformation failed, age worked. The Social Media Minimum Age Act passed in November 2024 with Coalition support and took effect on 10 December 2025, requiring platforms to take reasonable steps to keep under-16s off their services. On 8 September the government said more than five million underage accounts had been removed or deactivated and that more than 20 countries were following Australia's lead. Its own regulator's research told a more complicated story: in March–April 2026, eSafety found 81.5% of 10-to-15-year-olds still using at least one age-restricted platform, down from 85.9% before the ban began. The regime has kept expanding since. In June 2026 the government moved enforcement amendments lifting the age-ban penalties to about $99 million, compelling internal documents, and placing five platforms under investigation. A separate hate-speech and visa track was added in December 2025\. At Senate estimates on 27 May 2026 the department told senators the duty was not a misinformation revival. That is the government's through-line: different object, different words, adults facing only listed illegalities and a feed choice. Its opponents draw the opposite lesson from the same history — same government, same regulator, now holding an open harm list. ### **5\. The eSafety Commissioner's Record — and the Case It Lost** Because the duty hands so much to the regulator's judgement, the record of the office that would enforce it is the tell. The eSafety Commissioner would demand the risk assessments, set the rules for researcher data access, run safety tests using false accounts, and — in Wells' own words at the launch — "make that determination" on where harassment crosses the line. That office's recent history is mixed, and one loss in particular hangs over the new duty. After a 16-year-old stabbed Bishop Mar Mari Emmanuel during a livestreamed sermon at Christ the Good Shepherd Church in Wakeley in April 2024, eSafety sought a *global* takedown of the video. X geo-blocked it in Australia but refused a worldwide order, the Federal Court fight was dropped, and Musk branded Commissioner Julie Inman Grant a "censorship commissar." The Baumgarten case cuts closer. Celine Baumgarten posted a 53-second video criticising a "Queer Club" for Years 3–6 at Montmorency South Primary School. eSafety's own investigator found the post did not meet the legal test for cyber-abuse, because it was not intended to cause serious harm. The office sent X a complaint alert anyway; X geo-blocked the video in Australia; and when the matter reached court, the Full Federal Court dismissed eSafety's appeal on 18 February 2026, holding that an informal alert which produced a coercive geo-block was not beyond legal review. Taxpayer costs ran into the hundreds of thousands. Two more markers sit beside those. In July 2026 Inman Grant told the Royal Commission on Antisemitism and Social Cohesion that platforms were "fighting for the right and the ability to distribute and monetise gore and fringe content." In May 2026 the Federal Court fined X $650,000 over an unanswered 2023 transparency notice. The regulator that would police the new duty is the same one that lost Baumgarten. ### **6\. Censorship Warnings From the Coalition, One Nation and the IPA** The Coalition spent the launch week on the minister's rule-making power, not the feed prompt. Opposition Leader Angus Taylor told News24 on 6 September he was "deeply sceptical" and feared "this is just going to be an attempt by the government to censor social media," adding that "given Labor and the minister's past biases," he expected "an attempt to censor the internet." Deputy Liberal Leader Jane Hume said the Coalition had a "natural suspicion" of a minister seeking new powers to "control the internet," and put the objection plainly: "We don't feel that the minister should have the powers to determine what is and is not online harm." Nationals leader Matt Canavan argued Labor could not be trusted with the power at all — "we should never give censorship powers to people who are proven liars" — and cast the bill as displacement, saying the government, having "given up" on prices, tax and migration, now wanted "to censor the internet so you can't complain about these things." Pauline Hanson, the One Nation leader, went furthest. She said Labor was "copying the Chinese Communist Party's censorship playbook" — the lesson being that "you don't need to censor every citizen, you just need to create a culture of fear and make the platforms do it for you" — and warned the powers would "make sure social media companies feel the only option is to over-censor." She objected that "the minister and eSafety Commissioner get open-ended power to classify anything as harmful," and tied the timing to Labor's loss of the safe seat of Secret Harbour to One Nation at a Western Australian by-election, arguing the government was "desperately trying to censor the internet to keep their jobs." She also seized on the government's own language: "Today that's harm to children's safety. But tomorrow these powers will be abused to censor complaints about immigration because they harm Labor's version of 'social cohesion'. Social cohesion under Labor means sit down, shut up and don't complain." The free-market Institute of Public Affairs made the over-compliance case in colder terms, calling the harms vague, treating the package as "misinformation law by the back door," and coining "Alborithm" for a government-set feed regime. The government's own words from the same week gave the critics their material. In a Saturday Paper interview published on 5 September, Wells described constituents at her mobile offices who had read something online and then would not believe a cabinet minister to her face: "The power of the algorithm is so strong that it is overwhelming our public institutions, and the decline of trust is problematic for everybody." At the 8 September press conference she called that "a very separate question" from the duty. And asked about online radicalisation and foreign interference, Albanese said the National Security Committee of Cabinet had met the day before, and that among the concerns was "bad faith actors pushing material in order to undermine social cohesion. There is no question that that has occurred." ### **7\. Mission Creep: How the Duty Could Expand Without a New Act** The most important thing about the draft is not any harm currently on its lists. It is that the lists are open, and can be widened without Parliament passing another Act. By instrument, a minister could add a harm category on the "may cause serious harm" threshold; add a new design feature to the "harmful" list; tighten the user-control mandate from a two-option prompt toward, say, a chronological default or added friction; load extra severity metrics into the risk assessments; and refuse exemptions. None of that requires a fresh prohibition or a fresh vote on primary legislation. The only brake is Senate disallowance — which, on the YouTube precedent, has not been applied. The enforcement design pulls the same way. A government figure of $109.2 million per breach, stacked on the age-ban exposure and global reputational risk, is the kind of pressure under which a rational platform removes contested material first and argues afterwards. Baumgarten is the small-speaker version of that arithmetic: a geo-block first, a court vindication long after, and only for someone with the means to litigate. The draft carries no express free-expression duty, and no closed harm list. After the named offences are recited, what remains are the residual "anything else that could seriously harm a child" limb and the ministerial add-on powers. What those openings are eventually used for is not settled by the text. That they exist is. ### **8\. What Happens Next: Consultation, Timing and the Senate** Targeted consultation on the 7 September exposure draft is open until 12pm on 22 September (digitaldutyofcare@communications.gov.au), with introduction to Parliament promised later in 2026\. Commencement is split: if the bill passes and receives Royal Assent — the final step that turns an Act into law — the nudify notices, the 24-hour takedowns and search-engine link deletion would begin the next day, while the duty itself and the repeal of the industry codes would begin 12 months later. For now the government holds the centre. The Coalition is sceptical of the power to add harms by rule; the Greens want a harder opt-in on the feed. Albanese said the government was confident of the bill's constitutionality "or it would not be introducing" it, while a High Court challenge to the under-16 ban, brought by the civil-liberties group the Digital Freedom Project, is already on foot. Age-ban diplomacy continues around the United Nations gathering in New York. ### **Conclusion** The Digital Duty of Care draft would turn Australia's online-safety expectations into a fineable duty, price a single breach in nine-figure government figures, and present a two-option feed as user control. The same text lets a minister add harms and design features by rule and gives the eSafety Commissioner the last word at the harassment edge. The faster notice schedule would start the day after Assent; the duty itself a year later. What remains open is the consultation, the shape of the bill that finally reaches Parliament, and whether the Coalition or the Greens can move either the instrument power or the feed prompt. The government has ruled out a misinformation statute. It has not closed the power that a later rule could use to point the duty at speech.