Two governments are now arguing in public about whether an Australian rule on social media feeds is an American free speech problem, and the argument has a date on it. On September 22, the United States Embassy in Canberra filed a formal submission in Australia's consultation on the draft digital duty of care, saying it had serious concerns about provisions that would tell platforms how to rank the content people see. Australia's prime minister, Anthony Albanese, was in New York for the United Nations General Assembly when the submission was reported, and his answer was that the law is designed to hand people back control over their own devices rather than give the government control over what they read, according to NBC News.
The dispute is worth reading carefully, because it is not a disagreement about child safety. Both governments say they want children protected online. It is a disagreement about the instrument, and the instrument is the ranking system itself.
The objection is about who decides what reaches people
The embassy's submission makes three arguments, and they build on each other. The first is that the draft would let Australian regulators impose what the statement calls "rigid, one-size-fits-all platform design requirements," naming forced chronological feeds and opt-in recommendation systems as examples. The second is that the law's definitions of harm and risk are too vague to be applied consistently, which creates an incentive to over-moderate. The third is the conclusion: a regulator that can require a platform to filter what it recommends is doing something the United States treats as censorship.
The formulation the embassy used is worth stating exactly, because it is the hinge of the whole objection. A framework that lets a government enforce vague definitions of harm risks becoming, in the embassy's words, "a mechanism for viewpoint-based censorship," as RTÉ reported. The mechanism matters more than the intent. A rule written to protect teenagers from eating disorder content also gives a regulator a lever over ranking decisions, and once the lever exists, the question of who pulls it and against whom stops being hypothetical.
The submission then draws an equivalence that Australian officials have not yet answered directly. If a platform is compelled to cut the reach of content the law classifies as harmful, the embassy argued, reducing a speaker's audience is not different in kind from deleting the post. The United States asked Australia to withdraw any plan to mandate that platforms algorithmically suppress content based on ambiguous, government-defined safety criteria, as The Nightly reported.
What the draft requires of the platforms
Stripped of the framing, the Australian proposal has three moving parts, and only one of them is the algorithm rule that drew the objection.
The first is a choice about feeds. Platforms would have to notify users and offer a default feed of posts from accounts they follow, with the option to switch to algorithmically recommended content instead. The second is a duty of care toward users under 18, covering pornography, online bullying, content that glorifies crime and content that encourages eating disorders. Companies that fail would face fines of up to A$109.2 million. The third is enforcement machinery: the eSafety Commissioner would gain the power to issue removal notices to platforms.
The government opened the draft for consultation with digital platforms, industry bodies, civil society organizations and advocates, and says the bill will be introduced to parliament later this year. The scope of the consultation is the reason the embassy's submission matters as much as it does. A submission filed in a consultation is not a court filing and creates no legal obligation, but it is the form of objection that a drafting team has to answer on paper, and it is the record that a future dispute will be argued against.
The part of the design that decides how much the law changes anything is the one we wrote about when the draft was released: which feed the platform is allowed to make the default. Every regime that has offered a choice between an algorithm and a chronological timeline, from cookie banners to the European Union's Digital Services Act, has produced the same result, which is that most users keep the default. The embassy's submission and the government's defense are both about that setting, described in different vocabularies.
The extraterritorial argument is the strongest part of the American case
The embassy did not rest on principle alone. It made a jurisdictional point that Australia will find harder to dismiss: the large majority of the platforms affected are American companies, operating substantially on American soil, serving American users alongside Australian ones. A removal notice issued in Sydney would apply to a system that also serves people in Ohio, and the embassy said mandates of that kind could shape what people see and hear well beyond Australia's borders.
That is a claim about where the regulated conduct happens, and it is the argument the same embassy has been making about Australian platform regulation for over a year. It is also the argument that gives the American position whatever force it has. Australia can decide what its own regulators do. It cannot decide, on its own, that a compliance change made by a California company for Australian users will not alter the product its American users see, particularly when the platform runs one ranking system for everyone.
The Australian counter is structural rather than legal. Countries that regulate products sold by foreign companies accept that the rules travel with the market, and Australia has been explicit that platforms wanting Australian users accept Australian conditions. The difficulty is that speech is the one product where that exchange is contested, because the harm is not in the device but in the ranking, and the ranking is the same everywhere.
Canberra's answer is that choice is not control
Albanese's response, given to journalists in New York, was to reframe the measure as a transfer of power downward rather than upward. The prime minister said the reform does not expand the government's authority over what people read, and that he would defend the country's interests in the dispute, The Star reported. He made a fuller version of the same argument to a United Nations side event on children's rights in the age of artificial intelligence, appearing alongside President Emmanuel Macron of France and Prime Minister Pedro Sánchez of Spain, where he said of the new rules: "They will make social media platforms give people the choice."
The government's political foundation for this is the under-16 social media ban, which Australia implemented in December and which the prime minister describes as the base layer of the same project. That foundation is softer than the rhetoric suggests. Data reported last month found that under-16s in Australia were increasing their use of Instagram and TikTok despite the ban, which is the outcome that a compliance-based regime tends to produce when enforcement depends on self-reporting. Britain and France have moved in the same direction, so Australia is not alone, but it is the jurisdiction carrying the test case.
There is also a diplomatic signal in the timing. Albanese met Tim Cook at Apple's California headquarters the Sunday before the assembly, a one-hour closed-door meeting arranged at the chief executive's request after an earlier call about child safety features on Apple devices, some of which were modeled on the Australian ban. The government is negotiating with the companies it is regulating, while a third government argues in a consultation file that the regulation is unconstitutional in spirit.
Washington has used this word before, and the channel is new
The embassy submission is the third front in a pressure campaign that has been building since the middle of the year. In August, the White House described the News Bargaining Incentive, which charges platforms that fail to pay Australian news organizations, as a form of extortion, and in September the administration widened the charge to cover the duty of care itself. Kush Desai, a White House economics spokesman, said President Trump had warned trading partners against imposing "digital services taxes, fines, and other forms of extortion" on America's technology sector. In the same period, the National Foreign Trade Council attacked an Australian Tax Office ruling that treats cloud computing and streaming revenue as royalties, which would raise the tax bill for Amazon, Google, Microsoft, Apple, Spotify and Netflix.
What makes the consultation submission different from the rhetoric is that it arrives through a channel with an administrative function. A presidential statement is a signal. A formal submission in a rulemaking process is an input that the government must consider, summarize and respond to, and that later appears in the explanatory material attached to a bill. Officials can dismiss a spokesman's word choice. They cannot as easily dismiss a document filed by an allied government in their own consultation, particularly when the argument inside it is the one their own courts will eventually hear.
The American position has a softer edge than the language suggests. Trump did not raise the news bargaining charge in a recent call with Albanese, and reporting from Washington and the Australian embassy suggested the White House's attention to the technology file is lower than its attention to the trade disputes with China and Canada. The submission is the work of the machinery rather than the principal, which is exactly why it is durable.
What the consultation does not settle
Three questions will determine whether the law survives contact with the objection.
The first is the definition of harm. Australia has not yet published the standard it will use, and the embassy has asked it to specify how harm would be defined, who would decide, and against what benchmark. A narrow, enumerated list of harms is defensible. A general duty to reduce risk, enforced by a regulator with removal powers, is the version that turns a speech objection into a live legal argument.
The second is the default. If the government requires platforms to make the followed-accounts feed the default rather than an option, the rule bites. If it requires only that a choice be offered, the practical effect will be close to nothing, and the objection becomes easier to absorb without changing the substance of the law.
The third is enforcement reach. Removal notices against a platform that operates globally raise the extraterritorial question immediately, and Australia has not said how it would handle a notice that a platform declines to apply outside the country.
The bill goes to parliament later this year, which gives the government a short window to decide how much of the mandate it needs. One government has said that ranking decisions are design decisions, and that a country may regulate them. Another has said the same decisions are speech, and that regulating them is censorship. Both cannot be right about the same ranking system, and the drafting will show which one Australia decided to test.
Primary sources
- NBC News (Reuters), US warns Australia social media algorithm reforms risk censorship (September 23, 2026), for the embassy submission, its design-requirement and over-moderation arguments, the definition of harm request, and Albanese's response.
- RTÉ, Albanese defends Australia's draft social media laws after US criticism (September 23, 2026), for the A$109.2 million fine ceiling, the duty-of-care categories, the under-16 ban and its results, and the viewpoint-based censorship language.
- The Nightly, US criticises social media algorithm opt-out feature (September 23, 2026), for the withdrawal request, the eSafety removal notices, the US companies argument, the Tim Cook meeting, and Albanese's remarks to the UN coalition.
- The Star (AFP), Australia PM says draft social media rules about taking back control (September 23, 2026), for Albanese's remarks to journalists in New York.
- Brisbane Times, Trump spokesman cautions Australia over latest plan to punish Big Tech (September 9, 2026), for the White House extortion statement, the News Bargaining Incentive precedent and the Australian Tax Office royalty ruling.